國立交通大學 - national chiao tung universityeach sensing node is controlled by an imote2...
TRANSCRIPT
國立交通大學
土木工程學系
博士論文
整合式無線感測網路之結構健康監測系統
於建築與土木結構之研發
DEVELOPMENT OF AN INTEGRATED WIRELESS SENSOR
NETWORK-BASED STRUCTURAL HEALTH MONITORING
SYSTEM FOR BUILDINGS AND CIVIL
INFRASTRUCTURES
研 究 生:林 子 軒
指導教授: 洪 士 林 博士
中華民國一百年六月
整合式無線感測網路之結構健康監測系統
於建築與土木結構之研發
DEVELOPMENT OF AN INTEGRATED WIRELESS SENSOR
NETWORK-BASED STRUCTURAL HEALTH MONITORING
SYSTEM FOR BUILDINGS AND CIVIL INFRASTRUCTURES
研 究 生:林 子 軒 Student: Tzu-Hsuan Lin
指導教授:洪 士 林 博士 Advisor: Dr. Shih-Lin Hung
國 立 交 通 大 學
土 木 工 程 學 系
博 士 論 文
A Dissertation
Submitted to Department of Civil Engineering
College of Engineering
National Chiao Tung University
In partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy
in
Civil Engineering
June 2011
Hsinchu, Taiwan, Republic of China
中華民國一百年六月
I
整合式無線感測網路之結構健康監測系統
於建築與土木結構之研發
研究生:林子軒 指導教授:洪士林 博士
國立交通大學
土木工程學系
摘要
本研究提出一整合式無線感測網路之結構健康監測系統架構並應用於建築
與土木結構上。此架構主要包含了感測、資訊融合與管理以及決策診斷。在此架
構目標下,本研究發展了一個整合式無線感測網路結構健康監測系統。此系統由
多種無線節點所組成,包含感測結點、簇首節點、傳送節點與基地台。感測結點
主要以 Imote2 平台為基礎,用來量測結構的動態反應或是環境參數。簇首節點
為雙核心之設計結合了 Imote2 平台與額外之嵌入式系統。簇首節點含有額外之
無線通訊模組與 GPS來進行較長距離之資料接收、交換,並可以進行定位與同步
之用。傳送節點可當作協調者並進行資料的跳躍傳送(Hopping)。基地台主要為
接收所有回傳的感測資料與訊息,其硬體能力為所有結點中最高的。此系統亦包
含了一個三層式的軟體架構,主要可以進行可靠的資料感測與傳輸、資料儲存、
視覺化之使用者介面、資料分析與訊號處理等工作。在此軟體架構下,無線感測
相關的結構健康監測的應用將可以很容易地被實行。此外,能源的消耗亦是無線
感測的一個很重要的問題,因此,本研究發展了一個結合壓電、風車、與磁場的
整合式能源擷取系統來改善能源消耗的問題。 結構健康監測可分為局部與全域
之方法,然而,兩種方法各有其優缺點與應用之時機。因此本研究發展了一個整
II
合式的結構損害評估方法整合全域與局部之結構損害評估法,應用於結構健康監
測上。在全域結構損害偵測方面,本文分別提出了新的子結構頻率響應函數法來
對結構進行大範圍之損害評估。接著,以電機阻抗為主之局部損害評估法則可以
用來進行小範圍局部損害之評估。
在實驗與數值模擬研究中,數值模擬的結果發現本文所提出之結構損壞評估
方法可以成功地找出結構發生損壞的位置。另外無線感測網路結構健康監測系統
亦於一縮尺之建築模型上進行驗證,其實驗結果證實此系統可進行高質量的感測
與資料傳輸並可以精確地獲得結構的相關動態反應參數。為驗證所發展之無線感
測網路結構健康監測系統應用於實際土木結構上之可行性,將此系統佈設於一橋
梁上進行驗證。實驗結果顯示,此系統於實際土木之結構上依然可以獲得相當好
的感測結果,無線傳輸也可以達到預期的效果。在實際土木結構監測的佈設上更
可以發現本系統之優點,在實驗過程中,平均一個節點的建置小於 5分鐘,比起
傳統有線的監測系統來說,在時間成本的節省上更顯現出本系統的優勢。在六層
樓縮尺模型試驗中,亦可以證實本研究所提出之整合式的結構損害評估方法可以
識別出結構之全域與局部之損害。
III
DEVELOPMENT OF AN INTEGRATED WIRELESS SENSOR
NETWORKS-BASED STRUCTURAL HEALTH MONITORING
SYSTEM FOR BUILDINGS AND CIVIL INFRASTRUCTURES
Student: Tzu-Hsuan Lin Advisor: Dr. Shih-Lin Hung
Department of Civil Engineering
College of Engineering
National Chiao Tung University
ABSTRACT
The main purpose of this dissertation was to propose a framework for an
integrated wireless sensor network (WSN)-based structural health monitoring (SHM)
system in buildings and civil infrastructures. In this framework, three main parts were
considered: the physical sensing; information fusion and management; and inference
and decision making. To achieve this goal, an integrated WSN-based SHM system
was developed. This system consists of sensing nodes, cluster head nodes, transfer
node, and base station. The sensing node measures structural response or
environmental parameters. Each sensing node is controlled by an Imote2 platform
comprised of a microprocessor, sensor module, and communication device. To
exchange information or to trigger sensing tasks, the cluster head node can
communicate with sensing nodes or cluster heads of neighboring communities. The
cluster head has a dual-core design that combines the Imote2 platform with a second
embedded device. The cluster head has an extra wireless module and GPS. The extra
wireless module provides additional RF power needed for long-range wireless
IV
communication. The GPS is useful for synchronization and localization. The transfer
node functions as a coordinating node for managing cluster heads and data hopping.
The base station is the highest level end device and has the largest memory, the most
powerful processor and the highest communication capability. The base station node
is the gateway between smart sensor networks and the Host computer.
A three-tier software framework is also developed in this work serving as reliable
data-sensing and transmission, data logging and data storage, user interface, data
analyzing, and signal processing. Based on this software framework, a SHM
application for specific purpose can be easily developed. Power sources and power
consumption are the critical issue in WSN if batteries have to be periodically replaced.
Hence, a novel windmill-magnet integrated piezoelectric (WMIP) energy harvesting
system was also proposed.
Since local and global SHM have unique benefits and shortcomings, an
integrated approach may be more effective than using either approach alone. This
work developed a global-local-integrated damage detection approach for localizing
damage. Substructure-based frequency response function approaches were proposed
for global damage detection. Local damage was then identified by
Electro-Mechanical-Impedance (EMI)-Based damage detection method.
Numerical and experimental study is also conducted to complete this study.
Numerical results reveal that the proposed global damage detection approach can
successfully locate damage at a single site and at multiple sites. Experimental analysis
confirms the proposed integrated WSN-based SHM system provides excellent data
sensing and transmission quality for determining the structural dynamic properties.
V
An experimental validation in building structure confirms that the proposed global
SHM approach can indicate the approximate location of the damaged area in damaged
floor and the EMI-based damage detection approach can check the component of
building structure locally. Subsequently, the proposed WSN-based SHM system is
employed in a bridge structure to test the feasibility in field. This experimental result
confirms good-quality data collection by the proposed system. In experimental period,
the system shows WSN-based SHM system outperforms conventional SHM system,
especially in deploying sensors. Average time for deploying single node only takes
within 5min.
VI
ACKNOWLEDGEMENT
時間飛逝,轉眼間在交大也待了六個年頭了,學生生涯也在此畫下了句點。讀
博士真是一個漫長的過程,其中辛酸只有經歷過的人才能夠體會吧。在博士期間,
受到許許多多的幫助,在此獻上我最深的感謝。
當然最要感謝的是我的指導教授洪士林博士,他在研究上給予我非常大的自由,
讓我可以進行我有興趣的研究,在論文的方向與內容給予我相當大的幫助,並在
在其身上學到一個研究學者該有的風範。在生活方面,洪老師的國科會計畫也幫
助了我減輕經濟上的壓力。感謝黃炯憲老師在我論文中很重要的理論部分給予我
相當大的指導,在黃老師的身上也學到做研究嚴謹的態度。感謝廖德誠老師在電
機領域相關的指導,讓我跨領域的學習能夠如此的順利。另外,在博士班期間很
感謝東京大學的 Fujino 與 Nagayama 教授給我機會,讓我到東京大學訪問兩個
月,這兩個月讓我學習到很多東西,是個難忘的經驗。感謝靜宜大學吳仁彰教授
帶領我進入跨領域的研究,也感謝其時時為我禱告。感謝 Murthy 博士,無論是
在研究或是 paper的撰寫都給予我相當大的幫助,在他身上也學到很多。感謝中
興工程科技研究發展基金會,中興工程顧問社,中華顧問工程司,給予我的獎學
金。讓我能夠專心於研究上,不用為經濟擔心。
感謝研究室同學勇奇在我研究上有問題時可以互相討論,感謝研究室學弟們在
我實驗上的幫忙,感謝一弘在我修習電機相關課程時給予我幫助,並在 WSN的領
域一起努力學習,感謝阿銘在天線上面的幫助,感謝秦飛在 Imote2 上給予的協
助。還有許多在我博士班期間幫助過我的朋友,無法在此一一列出,在這一起獻
上我的感謝。也要感謝女友珮茹這幾年來對我的鼓勵與關心。最後要感謝我的家
人,沒有你們的支持,我沒有辦法專心地完成我的博士學位,感謝我的母親,您
辛苦了。
VII
CONTENTS
摘要 ............................................................................................................................................ I
ABSTRACT.................................................................................................................................. III
ACKNOWLEDGEMENT ...............................................................................................................VI
CONTENTS ............................................................................................................................... VII
LIST OF TABLES .......................................................................................................................... XI
LIST OF FIGURES ....................................................................................................................... XII
CHAPHER 1 INTRODUCTION ....................................................................................................... 1
1.1 BACKGROUND OF STRUCTURAL HEALTH MONITORING .................................................................. 1
1.2 MOTIVATION AND OBJECTIVE ................................................................................................... 3
1.3 OVERVIEW OF DISSERTATION .................................................................................................... 7
CHAPHER 2 LITERATURE REVIEW .............................................................................................. 10
2.1 INTRODUCTION ................................................................................................................... 10
2.2 STRUCTURAL HEALTH MONITORING APPROACH ......................................................................... 10
2.2.1 Global Damage Detection Method ........................................................................ 10
2.2.2 Local Damage Detection Method .......................................................................... 13
2.3 WIRELESS SENSOR NETWORKS-BASED STRUCTURAL HEALTH MONITORING APPLICATIONS ................. 15
CHAPHER 3 WIRELESS SENSOR NETWORKS .............................................................................. 18
VIII
3.1 INTRODUCTION TO WIRELESS SENSOR NETWORKS (WSN)........................................................... 18
3.2 SUBSYSTEM OF SMART WIRELESS SENSOR NODE ....................................................................... 20
3.3 WIRELESS SENSOR PLATFORMS............................................................................................... 21
3.4 FACTORS INFLUENCING SENSOR NETWORK DESIGN .................................................................... 23
3.5 THE 802.15.4 AND ZIGBEE STANDARDS .................................................................................. 25
3.6 OPERATING SYSTEMS FOR WIRELESS SENSOR NETWORKS ............................................................ 29
3.7 TIME SYNCHRONIZATION ....................................................................................................... 34
CHAPHER 4 INTEGRATED WSN-BASED SHM SYSTEM ................................................................. 37
4.1 INTRODUCTION ................................................................................................................... 37
4.2 ARCHITECTURE OF INTEGRATED WSN SHM SYSTEM .................................................................. 38
4.3 HARDWARE DESIGN ............................................................................................................. 39
4.3.1 Sensing Node ......................................................................................................... 40
4.3.2 Cluster Head and Transfer Node ............................................................................ 44
4.3.3 Base Station........................................................................................................... 47
4.3.4 Multi-Sensor Board Design .................................................................................... 48
4.3.5 Power Module ....................................................................................................... 52
4.4 WINDMILL-MAGNET INTEGRATED PIEZOELECTRIC (WMIP) ENERGY HARVESTING SYSTEM................. 52
4.4.1 General Theory of Vibration-Based Energy Harvesting Method ........................... 53
4.4.2 Energy Harvesting Circuits .................................................................................... 54
IX
4.4.3 Windmill-Magnet Integrated Piezoelectric Energy Harvesting System ................. 57
4.5 THREE-TIER SOFTWARE FRAMEWORK ...................................................................................... 62
4.5.1 Net Micro Frameworks-Based Embedded Development Environment .................. 64
4.5.2 Data Logging and Data Storage ............................................................................ 68
4.5.3 User Interface and Visualized Signal Processing ................................................... 69
4.6 RELIABLE DATA-SENSING AND TRANSMISSION SERVICE ................................................................ 72
CHAPHER 5 DEVELOPMENT OF GLOBAL-LOCAL-INTEGRATED DAMAGE DETECTION APPROACH76
5.1 INTRODUCTION ................................................................................................................... 76
5.2 SUBSTRUCTURE-BASED FREQUENCY RESPONSE FUNCTION FOR GLOBAL DAMAGE DETECTION ............ 78
5.2.1 Substructure-Based Frequency Response Function Approach ............................... 80
5.2.2 Numerical Study .................................................................................................... 91
5.2.3 Implement of SubFRFDI in Integrated WSN-Based SHM System ......................... 111
5.3 ELECTRO-MECHANICAL-IMPEDANCE (EMI)-BASED LOCAL DAMAGE DETECTION METHOD ............... 114
5.3.1 Fundamental of EMI-Based Model ...................................................................... 114
5.3.2 Measurement of Piezoelectric-Impedance .......................................................... 117
5.3.3 Selection of Frequency Ranges ............................................................................ 118
CHAPHER 6 EXPERIMENTAL VERIFICATION ............................................................................. 120
6.1 INTRODUCTION ................................................................................................................. 120
6.2 VERIFICATION OF INTEGRATED WSN-BASED SHM SYSTEM ........................................................ 120
X
6.2.1 Verification of Sensing Node ............................................................................... 120
6.2.2 Reliability of RF Communication Test .................................................................. 127
6.2.3 Field Test in Bridge Structure ............................................................................... 130
6.2.3.1 Experimental Setup ............................................................................................. 132
6.2.3.2 Experimental Results ........................................................................................... 134
6.3 EXPERIMENTAL STUDY IN BUILDING STRUCTURE ....................................................................... 140
CHAPHER 7 CONCLUSIONS REMARKS ..................................................................................... 158
7.1 CONCLUSION .................................................................................................................... 158
7.2 FUTURE STUDY .................................................................................................................. 161
REFERENCE ............................................................................................................................. 163
VITA ........................................................................................................................................ 176
XI
LIST OF TABLES
TABLE 3.1 DESCRIPTION OF KEYWORDS IN TINYOS. ............................................................................................ 33
TABLE 4.1 COMPARISON OF MOTE PLATFORM ................................................................................................... 43
TABLE 4.2 SPECIFICATION OF XBEE MODELS. ..................................................................................................... 46
TABLE 4.3 PHYSICAL AND ELECTRICAL CHARACTERISTICS OF GYRO SENSOR ............................................................. 50
TABLE 4.4 PHYSICAL AND ELECTRICAL CHARACTERISTICS OF ACCELEROMETER SENSOR .............................................. 51
TABLE 4.5 DIFFERENT TYPES OF PIEZOELECTRIC PRODUCT .................................................................................... 58
TABLE 4.6 DESCRIPTION OF CLASS LIBRARY ....................................................................................................... 66
TABLE 4.7 TOSMSG HEADER CONTENTS ........................................................................................................... 68
TABLE 4.8 SENSOR BOARD HEADER CONTENT .................................................................................................... 69
TABLE 4.9 DATA PAYLOAD .............................................................................................................................. 69
TABLE 5.1 DESCRIPTIONS OF SIMULATED DAMAGE CASES. .................................................................................... 92
TABLE 6.1 DESCRIPTION DAMAGE SCENARIOS FOR THE STEEL-FRAME MODEL. ........................................................ 144
TABLE 6.2 COMPARISONS OF MODAL PARAMETERS IDENTIFIED BY WIRELESS SENSING NODES AND CONVENTIONAL
REFERENCE SENSORS. .................................................................................................................. 151
TABLE 6.3 THE PROCESSING TIME OF EMBEDDED FFT AND FRF FOR DIFFERENT DATA LENGTH. .................................. 152
XII
LIST OF FIGURES
FIGURE 1.1 THE RELATION BETWEEN USAGE MONITORING, STRUCTURAL HEALTH MONITORING, AND DAMAGE PROGNOSIS. 3
FIGURE 1.2 A FRAMEWORK OF AN INTEGRATED WIRELESS SENSOR NETWORK-BASED STRUCTURAL HEALTH MONITORING
SYSTEM. ................................................................................................................................... 7
FIGURE 3.1 COMPOSITION OF WIRELESS SENSOR NODE ....................................................................................... 21
FIGURE 3.2 802.15.4 AND ZIGBEE ARCHITECTURE ............................................................................................ 26
FIGURE 3.3. STAR TOPOLOGY ......................................................................................................................... 27
FIGURE 3.4. PEER-TO-PEER TOPOLOGY ............................................................................................................ 28
FIGURE 3.5 SCHEMA OF TINYOS (CROSSBOW) .................................................................................................. 30
FIGURE 3.6 ARCHITECTURE OF TINYOS ( CROSSBOW) ........................................................................................ 31
FIGURE 3.7 THE PROCESS IN TINYOS( CROSSBOW) ............................................................................................ 32
FIGURE 4.1 INTEGRATED WSN-BASED SHM SYSTEM ARCHITECTURE: DIFFERENT ROLES ARE ASSIGNED TO NODES .......... 39
FIGURE 4.2 PHOTO OF SENSING NODE ............................................................................................................. 40
FIGURE 4.3 BLOCK DIAGRAM OF SENSING NODE ARCHITECTURE. ........................................................................... 42
FIGURE 4.4 BLOCK DIAGRAM OF CLUSTER HEAD NODE ARCHITECTURE. .................................................................. 46
FIGURE 4.5 PHOTOGRAPH OF CLUSTER HEAD NODE ON BOX ................................................................................ 47
FIGURE 4.6 BLOCK DIAGRAM OF BASE STATION ARCHITECTURE.............................................................................. 48
FIGURE 4.7 SIMPLIFIED BLOCK DIAGRAM OF MULTI-SENSOR BOARD ..................................................................... 49
FIGURE 4.8 PHOTOGRAPHY OF DEVELOPED MULTI-SENSOR BOARD........................................................................ 50
XIII
FIGURE 4.9 (A) CANTILEVER BEAM WITH TIP MASS, (B) MODEL OF EQUIVALENT LUMPED SPRING MASS SYSTEM .............. 54
FIGURE 4.10 TYPICAL PIEZOELECTRIC HARVESTING CIRCUIT .................................................................................. 55
FIGURE 4.11 ARCHITECTURE OF ENERGY HARVESTING BASED WIRELESS SENSOR NODE ............................................... 57
FIGURE 4.12 SCHEMATIC DIAGRAM OF WINDMILL BASED PIEZOELECTRIC HARVESTING SYSTEM .................................... 58
FIGURE 4.13 EXPERIMENT SETUP OF WINDMILL BASED PIEZOELECTRIC HARVESTING SYSTEM: (A) RAW VOLTURE, (B)MFC,
(C)PZT, (D) OVERPASS IN TOKYO UNIVERSITY. ............................................................................... 59
FIGURE 4.14 OUTPUT VOLTAGE OF PZT ........................................................................................................... 60
FIGURE 4.15 OUTPUT VOLTAGE OF MFC ......................................................................................................... 60
FIGURE 4.16 OUTPUT VOLTAGE OF RAW VOLTURE ............................................................................................. 61
FIGURE 4.17 WINDMILL-MAGNET INTEGRATED PIEZOELECTRIC (WMIP) ENERGY HARVESTING SYSTEM ........................ 61
FIGURE 4.18 OUTPUT VOLTAGE AND FFT OF WMIP ENERGY HARVESTING SYSTEM .................................................. 62
FIGURE 4.19 THE PROPOSED SOFTWARE ARCHITECTURE ...................................................................................... 63
FIGURE 4.20 STRUCTURE OF CLASS LIBRARY ..................................................................................................... 65
FIGURE 4.21. PACKET FORMAT ....................................................................................................................... 68
FIGURE 4.22 FUNCTIONS OF PROPOSED USER INTERFACE .................................................................................... 70
FIGURE 4.23 SENSOR DATA REPRESENTATION AND NODE DISTRIBUTION .................................................................. 71
FIGURE 4.24 SIGNAL PROCESSING INTERFACE .................................................................................................... 71
FIGURE 4.25 DAMAGE LOCATION REPRESENTATION ............................................................................................ 72
FIGURE 4.26. RELIABLE DATA-SENSING AND TRANSMISSION SERVICE ...................................................................... 75
XIV
FIGURE 5.1 GLOBAL-LOCAL-INTEGRATED DAMAGE DETECTION APPROACH ............................................................... 77
FIGURE 5.2 FREQUENCY RESPONSE FUNCTION MODEL ....................................................................................... 81
FIGURE 5.3. (A) ORIGINAL COMPLETE STRUCTURE, (B) THE ITH
SUBSTRUCTURE. ........................................................ 87
FIGURE 5.4. FRF OF SUBSTRUCTURE ............................................................................................................... 91
FIGURE 5.5 COMPARISONS OF CONVENTIONAL FRFS OF THE STRUCTURE IN UNDAMAGED AND DAMAGED STATES FOR CASE
1. ......................................................................................................................................... 95
FIGURE 5.6 COMPARISONS OF THE SUBSTRUCTURE-BASED FRFS FOR EACH SUBSTRUCTURE FOR CASE 1. ...................... 97
FIGURE 5.7 THE IDENTIFIED DAMAGE LOCATIONS IN SINGLE-DAMAGE CASES 1. ....................................................... 98
FIGURE 5.8 THE IDENTIFIED DAMAGE LOCATIONS IN SINGLE-DAMAGE CASES 2. ....................................................... 98
FIGURE 5.9 THE IDENTIFIED DAMAGE LOCATIONS IN SINGLE-DAMAGE CASES 3. ....................................................... 99
FIGURE 5.10 THE IDENTIFIED DAMAGE LOCATIONS IN SINGLE-DAMAGE CASES 4. ...................................................... 99
FIGURE 5.11 THE IDENTIFIED DAMAGE LOCATIONS IN SINGLE-DAMAGE CASES 1 FOR DIFFERENT DAMAGE EXTENT. ........ 100
FIGURE 5.12 THE IDENTIFIED DAMAGE LOCATIONS IN SINGLE-DAMAGE CASES 3 FOR DIFFERENT DAMAGE EXTENT. ........ 100
FIGURE 5.13 THE VALUES OF THE SUBFRFDI OF EACH SUBSTRUCTURE IN DAMAGE CASE 3, OBTAINED USING TWO
DIFFERENT FREQUENCY RANGES. ............................................................................................... 103
FIGURE 5.14. THE FRF OF SUBSTRUCTURE: WITH AND WITHOUT NOISE. ............................................................. 103
FIGURE 5.15 THE EFFECTS OF NOISE ON SUBFRFDII ESTIMATIONS FOR CASE 3. ..................................................... 104
FIGURE 5.16 THE SUBFRFDI VALUES OF DIFFERENT SUBSTRUCTURES IN DAMAGE CASE 4 WITH DIFFERENT EARTHQUAKE
EXCITATIONS. ......................................................................................................................... 104
XV
FIGURE 5.17 THE ANALYTICAL RESULTS FOR DAMAGE CASE 3 OBTAINED USING THE PROPOSED METHOD (SUBFRFDI) AND
THE FRF CURVATURE-METHOD-BASED INDEX (FRFCDI). ............................................................... 105
FIGURE 5.18 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 7, MULTIPLE DAMAGE. ........................ 106
FIGURE 5.19 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 8, MULTIPLE DAMAGE. ........................ 106
FIGURE 5.20. THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 9, MULTIPLE DAMAGE. ....................... 107
FIGURE 5.21 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 10, MULTIPLE DAMAGE. ...................... 107
FIGURE 5.22 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 11, MULTIPLE DAMAGE. ...................... 108
FIGURE 5.23 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 12, MULTIPLE DAMAGE. ...................... 108
FIGURE 5.24 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 13, MULTIPLE DAMAGE. ...................... 109
FIGURE 5.25 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 14, MULTIPLE DAMAGE WITH VARIOUS
DAMAGE LEVEL. ..................................................................................................................... 109
FIGURE 5.26 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 15, MULTIPLE DAMAGE WITH VARIOUS
DAMAGE LEVEL. ..................................................................................................................... 110
FIGURE 5.27 THE SUBFRFDI VALUES FOR SUBSTRUCTURES IN DAMAGE CASES 16, MULTIPLE DAMAGE WITH VARIOUS
DAMAGE LEVEL. ..................................................................................................................... 110
FIGURE 5.28 AN SUBFRFDI-BASED PROTOCOL ............................................................................................... 113
FIGURE 5.29. DIAGRAM OF PZT-STRUCTURE BONDED SYSTEM ........................................................................... 115
FIGURE 5.30. BLOCK OVERVIEW OF AD5933 ................................................................................................. 118
FIGURE 6.1 ACCELERATION RECORD FROM WIRELESS SENSING NODES AND REFERENCE SENSOR. ................................ 121
FIGURE 6.2 FFT OF ACCELERATION RECORD FROM WIRELESS SENSING NODES AND REFERENCE SENSOR. ...................... 122
XVI
FIGURE 6.3 POWER SPECTRAL DENSITY OF ACCELERATION RECORD FROM WIRELESS SENSING NODES AND REFERENCE SENSOR.
.......................................................................................................................................... 122
FIGURE 6.4 A 1/8-SCALED THREE-STOREY STEEL FRAME MODEL. ........................................................................ 124
FIGURE 6.5 ACCELERATION RECORD FROM WIRELESS SENSING NODES ................................................... 125
FIGURE 6.6. RESPONSE FOR ANN AND WIRELESS SENSOR NODE ......................................................................... 126
FIGURE 6.7 SPRR FOR VARIOUS PERIOD MEASUREMENT ................................................................................... 128
FIGURE 6.8 COMPARISON OF SUCCESS PACKETS RECEIVED RATE(SPRR) FOR WELL TIME SCHEDULE AND NON TIME
SCHEDULE. ........................................................................................................................... 129
FIGURE 6.9. SUCCESSFUL PACKET RECEIVED RATE (SPRR) TO RF POWER AND DISTANCE ......................................... 130
FIGURE 6.10. THE EXPERIMENT SETUP AND SENSOR LOCATIONS .......................................................................... 133
FIGURE 6.11 PREVIOUS TENSILE TEST ........................................................................................................... 133
FIGURE 6.12 PHOTO OF NIOUDOU BRIDGE .................................................................................................... 134
FIGURE 6.13 WEAK PART OF THE STEEL BAR ................................................................................................... 135
FIGURE 6.14 EXCAVATION OF SOIL GROUND, ................................................................................................... 135
FIGURE 6.15 WIRELESS SENSING NODE POSITION. ........................................................................................... 136
FIGURE 6.16 MODIFIED RELIABLE DATA-SENSING AND TRANSMISSION SERVICE ...................................................... 137
FIGURE 6.17 RESPONSES OF SENSOR NODES FOR INTACT CASE. ........................................................................... 138
FIGURE 6.18 COMPARISON OF SCOURING AND INTACT BRIDGE PIER WITH FREE VIBRATION RESPONSE ON THE TOP NODE.
.......................................................................................................................................... 139
XVII
FIGURE 6.19 COMPARISON OF FFT FOR SCOURING AND INTACT BRIDGE PIE. ......................................................... 139
FIGURE 6.20 EXPERIMENTAL SETUP FOR WIRELESS SHM SYSTEM AND IMPEDANCE MEASUREMENT DEVICE ................. 140
FIGURE 6.21 SUBFRFDI VALUE: DAMAGE ON FIRST FLOOR. ............................................................................... 141
FIGURE 6.22 A BASELINE OF UNDAMAGED STATE AND FOUR DAMAGED IMPEDANCE CURVES. .................................... 142
FIGURE 6.23 A DAMAGE METRIC CHART OF RMSD FOR DAMAGE ON JOINT 4. ...................................................... 143
FIGURE 6.24 A PHOTOGRAPH OF THE TEST STRUCTURE. (B) THE IMOTE2.NET-BASED SENSING NODES FIXED ON THE FLOOR.
.......................................................................................................................................... 144
FIGURE 6.25 PHOTOGRAPHER OF DAMAGE SCENARIO: (A) DAMAGE SCENARIO 1. (B) DAMAGE SCENARIO 2. (C) DAMAGE
SCENARIO 3. (D) DAMAGE SCENARIO 4. ...................................................................................... 145
FIGURE 6.26 ACCELERATION OF WIRELESS SENSOR AND REFERENCE SENSOR .......................................................... 147
FIGURE 6.27. FRF OF THE STRUCTURE: (A)1F, (B)2F, (C)3F, (D)4F, (E)5F, (D)6F .................................................. 150
FIGURE 6.28 COMPARISONS OF MODAL PARAMETERS IDENTIFIED BY EMBEDDED FRF AND ANNSI. ........................... 152
FIGURE 6.29 THE SUBFRFDI FOR DAMAGE SCENARIOS 1–4. ............................................................................. 154
FIGURE 6.30 PZT SETUP ON THE TEST STRUCTURE. .......................................................................................... 155
FIGURE 6.31 RMSD FOR DAMAGE SCENARIO1 ............................................................................................... 156
FIGURE 6.32 RMSD FOR DAMAGE SCENARIO2 ............................................................................................... 156
FIGURE 6.33 RMSD FOR DAMAGE SCENARIO3 ............................................................................................... 157
FIGURE 6.34 RMSD FOR THREE DAMAGE SCENARIOS ....................................................................................... 157
1
CHAPHER 1 INTRODUCTION
1.1 Background of Structural Health Monitoring
Monitoring the structural health of buildings and civil infrastructure has received
considerable interest recently. A low-cost, real-time monitoring system with high
stability and robustness is required. A successfully implemented structural health
monitoring (SHM) system has several benefits to the end user of a structural system.
For example, a well designed SHM can reduce risk to operators and structural system,
reduce risk through life-extending operation and control based on the use of health
monitoring information and reduce cost for servicing structural systems based on
condition-based maintenance scheduling [1].
Monitoring the structural health of a given structural system is a damage
identification process that includes damage detection, damage localization, damage
type evaluation, and damage severity estimation. Damage can be defined as changes
to a structural system, such as its material and/or geometric properties, that alter its
current or future performance [2, 3]. Structural health monitoring can be classified as
local and global monitoring. Non-destructive evaluation (NDE) techniques are the
most widely used methods for local health monitoring. Conventionally adopted global
structural monitoring methods are vibration-based schemes. These methods identify
damage by detecting modal property changes in, say, natural frequencies, modal
damping, or mode shape [4].
The SHM problem is fundamentally one of statistical pattern recognition
paradigm. This paradigm can be described as a four-part process [5]:
2
(1). Operational Evaluation:
Define damage for the system being monitored, realize the conditions, both
operational and environmental, under which the system to be monitored, know the
limitations on acquiring data in the operational environment.
(2). Data Acquisition, Fusion and Cleansing:
The data acquisition portion of the SHM process involves selecting the types of
sensors, choosing the locations where the sensors should be placed, deciding the
number of sensors to be used, and selecting the types of data
acquisition/storage/transmittal hardware.
(3). Feature Extraction and Information Condensation:
Feature extraction is the process of the identifying damage-sensitive properties,
derived from the measured system response, which allows one to distinguish between
the undamaged and damaged structure.
(4). Statistical Model Development for Feature Discrimination.
The algorithms used in statistical model development usually fall into t fall into
the general classification referred to as supervised learning, unsupervised learning,
group classification and regression analysis.
After SHM was established, the next stage is a lead-in of damage prognosis. The
damage prognosis is a process of predicting the future probable capability of a
structural material or system in an online manner, taking into account the effects of
damage accumulation and estimated future loading. Figure 1.1 illustrates the relation
between usage monitoring, structural health monitoring, and damage prognosis [1, 6].
3
Figure 1.1 The relation between usage monitoring, structural health monitoring,
and damage prognosis.
1.2 Motivation and Objective
In SHM, integration of cross discipline is essential for developing a flexible and
robust system. For example, Civionics is currently the use of electronics for structural
health monitoring (SHM) of civil structures. It is combination of electronic
engineering with civil engineering, in a manner similar to avionics (aviation and
electronics) and mechatronics (mechanical engineering and electronics) [7].
Generally, sensors comprise a significant portion of the SHM process. Recent
developments in smart sensor technology enable new applications in structural health
monitoring. The main features of a typical smart sensor are on-board microprocessor,
sensing capability, data storage, wireless communication, battery power, and low cost.
Actual Loading
and operating
condition
Structural
system
Input and
environment
measurement
Response
measurement
Usage Monitoring
System Assessment Model
Structural Health Monitoring
Damage Accumulation
Model
Damage Prognosis
Predictive Loading Model
Future Loading Model
Design Loading
4
Sensors are being deployed in civil infrastructures. Long term recorded data for
monitoring are extensive. Smart sensors can process data before outputs are recorded,
which reduces the quantity of data required and computing power [8].
When many sensors are implemented, wireless communication appears to be an
attractive approach. Wired sensor systems can only deploy limited numbers of sensors
because of cost constraints or excessive complexity. Wireless sensors are expected to
minimize these problems by simplifying the installation of wired sensors [9-11].
Smart sensor-based wireless sensor networks (WSNs) are an attractive sensing
technology for structural health monitoring applications because of their low
manufacture costs, low power requirements, small size, and simple deployment (i.e.,
lack of cables) [12, 13].
Although numerous researchers have demonstrated the advantages of WSN
[12-17], there are still existing many limitations. First, the wireless throughput is
heavy to collect all of the measured information in SHM. The aggregation of dynamic
measurement data consumes power- and time- consuming even when data are not
collected in real time. For instance, Kim et al. [18] developed a multi-hop wireless
sensor network to monitor the Golden Gate Bridge. They reported that transferring
KB data from 64 nodes required over 12 hours. Moreover, sensor nodes are not
synchronized with separate clocks. Packets may be lost in communication or sensing,
and storage memory space is always limited. Therefore, an effective time-scheduling
and data transmission protocol should be developed.
Additional, the processing ability of processor of node in WSN is slower than
that of a PC. In SHM, a complex parameter identification-based embedded damage
detection approach may not perform adequately when using resource-constrained
5
WSN hardware. Therefore, a non-parameter embedded damage detection approach
such may be considered a appreciate approach.
Next, battery powered wireless nodes have limitations on consuming large
amounts of power in packets delivering. Hence, power sources and power
consumption are the critical issue if batteries have to be periodically replaced.
Therefore, an intelligent self -powered wireless sensor networks via harvest and store
ambient sources of energy should be considered to solve this barrier.
Furthermore, implementing an embedded algorithm in a complex development
environment commonly involves hardware and software complexities, making the
task quite challenging, especially for civil engineers with limited experience of WSN.
Also, robust system development is problematic in a complex programming
environment. Accordingly, a WSN-based SHM system should be developed based on
an easy-to-use development environment. Related applications can be implemented
efficiently in this the friendly environment.
The main purpose of this dissertation is to propose a framework of an integrated
wireless sensor network-based structural health monitoring system in buildings and
civil infrastructures. Figure 1.2 illustrates the proposed framework. In this framework,
it is considered three main parts which are the physical sensing, information fusion
and management, and inference and decision making. To achieve this goal, an
integrated WSN-based SHM system is developed. This system consists of sensing
nodes, cluster head nodes, transfer node, and base station. The purpose of sensing
node is to measure the responses of structure or the environmental parameters. Cluster
head node can communicate with the sensing nodes or other cluster heads of the
neighboring communities to exchange information or triggering sensing task. The
6
transfer node is functions as coordinate node for managing the cluster heads and
hoping the data. The base station is the highest level end device that has largest
memory, most powerful processor and highest communication capability. The base
station node is the gateway between smart sensor networks and the Host computer.
The base station uploads the data to remote monitoring and control server via satellite,
3G telecommunication or WiMAX communication. A three-tier software framework
is also developed in this work serving as reliable data-sensing and transmission, data
logging and data storage, user interface, data analyzing, and signal processing. Based
on this software framework, a SHM application for specific purpose can be easily
developed. Power sources and power consumption are the critical issue in WSN if
batteries have to be periodically replaced. Hence, a novel windmill-magnet integrated
piezoelectric (WMIP) energy harvesting system was also proposed.
Both local and global SHM each has its own benefits and shortcoming, hence
integrating aforesaid two approaches can obtain more effective damage detection
result than only using one of them. This work proposed a global-local-integrated
damage detection approach to localize damage. Substructure-based frequency
response function approaches are proposed as global damage detection approach.
Electro-Mechanical-Impedance (EMI)-Based damage detection method is used to
identify local damage of structure in this study. In addition to theoretical
developments in damage assessment approach and developing of wireless SHM
system, experimental study is also conducted to complete this thesis.
7
Figure 1.2 A framework of an integrated wireless sensor network-based
structural health monitoring system.
1.3 Overview of Dissertation
This research focuses on realization of a vibration based SHM framework
employing WSN based SHM system that can detect and localize damage. In
addition to introduction in Chapter 1, Chapter 2 provides a review related to damage
detection method and WSN-based SHM approaches that have been developed in
literatures.
Since the WSN plays a significant role in this study, its basic theories and related
techniques are briefly introduced in Chapter 3. In this Chapter, the basic wireless
sensor technology would be introduced. WSN based architecture and protocol stack
and operating systems for WSN then addressed. Time synchronization and related
Integrated WSN-based SHM system
Data Representation
Viewer
Sensor Nodes Topology
Viewer and Health
Monitoring
Signal Processing
and Analysis
Structural Health
Monitoring and Warning
Multi-Sensors
Structure
Data Fusion
Sensor Control and Management
Multi-Damage Detection Approach Inference
Physical SensingInformation Fusion
and Management
Inference and
Decision Making
Global-Local-Integrated Damage Detection Approach
Substructure-based FRF
approaches
Electro-Mechanical-Impedance (EMI)-Based
damage detection method
Structural
Health
Monitoring
and Warning
8
applications are also included in this Chapter.
After the theoretical basis of WSN is established, the proposed integrated
WSN-based SHM system then presented in Chapter 4. The architecture of this SHM
system and wireless sensor nodes design are presented in Chapter 4. Chapter 4 also
shows the proposed Windmill-magnet integrated piezoelectric (WMIP) energy
harvesting system which is used to harvesting ambient energy. The software is also an
important part in this system. The proposed three-tier software framework and reliable
data-sensing and transmission service are hence included in this chapter.
Chapter 5 shows the proposed global-local-integrated damage detection approach.
The theoretical basis and numerical study are presented in this chapter. The
substructure-based frequency response function based global damage detection
approaches are presented. Subsequently, the Electro-Mechanical-Impedance
(EMI)-based local damage detection method is proposed herein to detect the local
damage.
In Chapter 6, this dissertation focused on the experimental study. When the
theoretical basis of damage detection method and wireless SHM system based
measuring system are completed, series experimental study are conducted to verify
the proposed algorithm and system. First off all, the wireless SHM system is verified
involving sensor calibration, wireless communication quality, data losing, interference
in material, power consumption. Subsequently, the proposed wireless SHM system is
employed in a bridge structure to test the feasibility in field. Next, the proposed
damage detection approach with proposed wireless SHM system then investigated in
a ¼ -scale six-storey steel structure that was designed by the National Center for
Research on Earthquake Engineering (NCREE), Taiwan.
9
In the end of this dissertation, Chapter 7 summarizes the research detailed on the
analytical and experimental study in this dissertation and presents possible directions
for future research on SHM using wireless sensor networks.
10
CHAPHER 2 LITERATURE REVIEW
2.1 Introduction
This chapter provides a review related to SHM techniques and WSN-based SHM
approaches that have been developed in literatures. Since the SHM can be classified
as global and local monitoring, the global and local damage detection methods are
briefly reviewed respectively. Moreover, this work focus on using WSN related
techniques in developing SHM system. Some representative applications of
WSN-based SHM, including theoretical developments, experimental validations, and
practical applications, will be briefly reviewed in this chapter.
2.2 Structural Health Monitoring Approach
2.2.1 Global Damage Detection Method
Some global health monitoring methods are studied on either finding shifts in
resonant frequencies or changes in structural mode shapes. Early works in health
monitoring found that loss of a single member in a structure can result in changes in
the fundamental natural frequency of one to as much as thirty percent. However, it is
easy to see that some damage forms may not affect the frequency at low levels of
vibration [8]. Although locating damage using frequency shifts requires either very
accurate measurements or large damage extent, recent researches have shown that
using resonant frequencies can get much less statistical variation from random error
sources than other modal parameters [19, 20]. More detailed discussion related to
damage detection using shifts resonant frequencies can be found in [21, 22]. West [23]
11
presented a efficient method using mode shape information for locating structural
damage without the accurate FEM model. Mayes [24] developed a structural
translational and rotational error checking (STRECH) method for locating model error
by mode shape changes. Other studies provide instances of those mode shapes-based
modal assurance criteria (MAC) and coordinate MAC (COMAC) values can be used
to identify damage [25-27].
Early researchers [28] indicated that mode shape changes are not sensitive to
locate local damage, an improved method for locating damage is to use the mode
shape curvature. The using of mode shape curvature appears to be more sensitive for
detecting damage due to losing stiffness of member than the mode shapes themselves
[29]. Pandey [29] using the central difference operator to compute curvature values
from the displacement mode shape. Their results demonstrate that changes in mode
shape curvature can be a good damage index for the FEM beam structures. Topole and
Stubbs [30] survey the practicality of using a limited set of modal parameters for
detecting structural damage. Chance et al. [31] used measured strains instead to
measure curvature directly, which dramatically improved unacceptable errors in
calculating curvature from mode shapes.
Another class of damage detection method is based on the structural model
matrices updating such as mass, stiffness, and damping. By solving the constrained
optimization problem, this method updates model matrices using the structural
equations of motion, the nominal model, and the measured data. Based on the various
algorithms, the method can be classified as optimal matrix update methods [32-35],
sensitivity-based update methods [36-39] and eigenstructure assignment method [40,
41]. These methods have some limitations such as the baseline of stiffness, mass, and
damping matrices are probably inaccurate, and the solution of the optimization is not
12
unique [8].
Statistical pattern-recognition approaches are widely used in Global SHM. Sohn
et al. [42] presented two pattern recognition techniques using time series method to
analyze fiber optic strain gauge data obtained from two different structural conditions
of a surface-effect fast patrol boat. The first technique combines Auto-Regressive (AR)
and Auto-Regressive with eXogenous inputs (ARX) prediction models and the second
technique employs outlier Mahalanobis distance measure-based analysis. These two
methods were successfully employed to identify different structural damages by using
data sets changes. The method in [42] is further applied to the damage detection of the
benchmark problem designed by the ASCE task group [43]. The damage detection
approach on the originally extracted damage-detection feature is modified to consider
the influence of excitation variability and the orders of the ARX prediction model.
Medium and severe damages are successfully detected and localized. This approach is
investigated using various acceleration responses generated with different
combinations of structural finite element models, excitation conditions and damage
patterns in the benchmark study.
Artificial neural networks (ANNs) have been recently studied in SHM. Chen et
al. [44] used ANNs to identify the structural dynamic parameters by using the
structural responses in a building structure subjected to earthquake. The simulation
results showed that the structural dynamic behavior can be well modeled by the
trained neural networks. Wu et al. [45] employed ANNs for detecting structural
damage in a three-story frame. The ANNs was used to recognize the frequency
response characteristics of undamaged and damaged structures. The changing damage
extents were simulated by changing the properties of individual members. Wen et al.
[46] presented an ANNs-based approach for detecting structural damage. This work
13
adopts an unsupervised neural network which incorporates the unsupervised fuzzy
neural network (UFN) to localized damage. The damage localization feature (DLF) is
applied to locate structural damage by using the UFN. The analysis results indicated
that the use of fuzzy relationship in UFN made the detection of structural damage
more robust and flexible than the back-propagation network (BPN). In [47], ANNs
has been used to training a set of strain measurements for different types of damage in
a structure. Although ANNs can identify new damage type and location patterns, with
a limited set of training examples, convergence is not guaranteed.
In addition to above method, wavelet-based damage detection method is also
widely used in SHM [48-50]. The advantage of using wavelets is the capability to
perform local analysis of a signal. Hence, wavelet analysis can find hidden
information in data that other signal analysis methods may fail to detect. This property
is essential for effective damage detection [50]. The wavelet-based damage detection
approach is widely used to identify modal parameters of structural systems [51, 52],
to detect beam structure damage [49, 53, 54], and to de-noise of signals [55, 56]. The
discrete wavelet transform (DWT) shows that the de-noising capability is sufficient
for detecting drill fractures and bearing race faults [56-58].
2.2.2 Local Damage Detection Method
The Non Destructive Testing (NDT) technique is mostly used in local damage
detection. There are many methods for NDT such as ultrasonic, infrared thermography,
eddy current X–ray. Such methods can detect tiny or incipient damage; however, it is
usually expensive, power hungry and bulky. Sansalone and Carino [59] developed an
impact-echo method for flaw detection in relatively thin concrete structures. They
indicated that use the impact-echo method with frequency analysis can successfully
14
detect flaw in thin concrete structures.
Another impact method is known as impulse response, transient dynamic
response, or impedance testing [60]. In their approach, both force history of the
impact and the response of the structure are measured. The typical impulse response
spectrum of the structure can be acquired using measured response and force history
by using a signal processing technique. The impulse response spectrum of a structure
reflects to the characteristic of structure such as geometry, the support conditions, and
the cracks on structure.
The use of eddy current method has been developed for detecting
surface-breaking cracks in the area of welded connections. One advantage of this
method is that it can detect surface cracks of steel without removing paint.
Commercial sensors have been designed for detecting the cracks of welds in highway
bridges in the field. This eddy current-based method has successfully been employed
in detecting the crack [61, 62].
Ultrasonic time of flight diffraction (TOFD) method has been developed to
detecting and quantifying eye bars cracking. Compare to pulse-echo methods used the
specular reflections from the crack face, this method is based on using the wave
diffraction from a crack tip. Washer [63] applied this technology in the detection of
eye bar cracks.
Electro-Mechanical-Impedance (EMI)-Based damage detection method is a new
method of performing NDT by analyzing the electromechanical coupling property of
piezoelectric materials. The EMI-based damage detection approach uses active
surface-bonded PZT (Lead zirconate titanate) patches to sense structural-mechanical
impedance changes. Liang et al. [64] first proposed an electrical admittance model of
15
the a PZT bar connected to a structure. Sun et al. [65] then experimentally used
electrical impedance method to acquire FRF for a single location and to transfer FRF
between two locations on a structure. This technique has proven effective in several
experimental applications, including aircraft structures [66], temperature-variable
applications [67], civil structural components [68], concrete structures [69], and PSC
girder bridges [70]. Impedance methods also continue to attract the attention of
researchers and field engineers. Park et al. [71] summarized future issues of the
impedance method from hardware and software standpoints. Such issues include the
difficulties in handling the crisp PZT sensors and in bonding them to the structure, the
bulky and expensive analyzers required for impedance testing, the difficulties and
complexities of acquiring and processing data in large-scale complex structures, and
high power consumption.
2.3 Wireless Sensor Networks-Based Structural Health Monitoring
Applications
Recently, smart sensor-based WSN has been considered as an alternative
technology for SHM. In this section, available wireless sensor platforms developed in
the academia and industries are reviewed, and applications of WSN to SHM of civil
structures are also reviewed.
Mascarenas et al. [72] developed a mobile host-based WSN monitoring system.
The mobile host transmitted energy to power the sensor node wirelessly. When the
sensor node has received adequate energy for sensing, the sensor node then start
sensing and wirelessly transmits the measurement to the mobile host. They used a
commercially available radio-controlled helicopter to deliver microwave energy to
wireless sensor nodes located on a decommissioned overpass in southern New
16
Mexico. The helicopter successfully delivered sufficient microwave energy to charge
a wireless sensor node to get displacement measurements and transmit the data back
to the helicopter. The results of this experiment show that a mobile-based wireless
sensor network can feasibly be used to SHM.
Kim et al. [73] proposed a WSN SHM system which was deployed and tested on
the 4200ft long main span and the south tower of the Golden Gate Bridge (GGB). 64
nodes are distributed deployed in the GGB deployment, over the main span and the
tower, collecting ambient vibrations synchronously at 1kHz rate, with less than 10μs
jitter, and with an accuracy of 30μG. The sampled data is collected via a 46-hop
wireless network, with a bandwidth of 441B/s at the 46th hop. The collected data
agrees with theoretical models and previous studies of the bridge. The deployment is
the largest WSN for SHM.
Cho et al. [74] presented a structural health monitoring (SHM) system using a
dense array of scalable smart wireless sensor network on a cable-stayed bridge (Jindo
Bridge) in Korea. 70 sensors and two base station computers have been deployed to
monitor the bridge. This autonomous SHM application is consideration with harsh
outdoor surroundings. The performance of the system has been tested in terms of
hardware stability, software reliability, and power consumption. 3-D modal properties
were extracted from the measured 3-axis vibration data using output-only modal
identification methods.
Potential of WSNs for using in bridge management has been evaluated in [75].In
their study, a network of seven sensor nodes was installed on the Ferriby Road Bridge,
a three-span reinforced concrete bridge. Three displacement sensor nodes were placed
across cracks on the soffit of the bridge to measure the change in crack width. Three
17
inclinometer sensor nodes were mounted on two of the elastomeric bearing pads to
measure the change in inclination of the bearing pads. Another node was used to
monitored temperature in the box that contained the gateway. They suggest that the
sensors are stable enough for long-term use. Sensor networks represent a useful tool
that can be used to supplement, but not replace, visual inspection.
Recently, a mobile wireless sensor network is proposed for installation on a
heavy truck to capture the vertical acceleration, horizontal acceleration and
gyroscopic pitching of the truck as it crosses a bridge [76]. The vehicle-based wireless
monitoring system is designed to interact with a static, permanent wireless monitoring
system installed on the bridge. Specifically, the mobile wireless sensors
time-synchronize with the bridge's wireless sensors before transferring the vehicle
response data. Vertical acceleration and gyroscopic pitching measurements of the
vehicle are combined with bridge accelerations to create a time-synchronized vehicle–
bridge response dataset. In addition to observing the vehicle vibrations, Kalman
filtering is adopted to accurately track the vehicle position using the measured
horizontal acceleration of the vehicle and positioning information derived from
piezoelectric strip sensors installed on the bridge deck as part of the bridge monitoring
system. Using the Geumdang Bridge (Korea), extensive field testing of the proposed
vehicle–bridge wireless monitoring system is conducted. Experimental results verify
the reliability of the wireless system and the accuracy of the vehicle positioning
algorithm.
18
CHAPHER 3 WIRELESS SENSOR NETWORKS
3.1 Introduction to Wireless Sensor Networks (WSN)
Recently, advanced micro-electro-mechanical systems (MEMS) technology,
wireless communications, and electronics engineering have improved the
development of WSN. WSN is based on wireless sensor nodes. These tiny wireless
sensor nodes, which consist of sensing, processing, and communicating components,
collaborate with each other via wireless communication.
WSN have great potential for many applications in different scenarios such as
natural disaster monitoring [77], military target tracking and surveillance [78],
biomedical health care [14], and hazardous environment investigation and seismic
sensing [79]. In military target tracking and surveillance, a WSN can assist army to
detect and identify enemy target such tank movements as spatially-correlated and
coordinated troop. With natural disasters, wireless sensor nodes were used to sense
and detect the environment to forecast disasters before they occur. In biomedical care
applications, the wireless sensors can help monitor a patient’s health in surgical. For
seismic sensing, deployment of large numbers of wireless sensors along the volcanic
area can detect the development of earthquakes and eruptions.
A wireless ad hoc network is a kind of wireless technology. It is decentralized
type of wireless network. The network does not rely on a preexisting infrastructure.
Instead, each node forward data dynamically based on the network connectivity. Ad
hoc networks can use flooding for forwarding the data. There are differences between
WSN ad hoc networks. The differences between WSN and ad hoc networks was
19
illustrate below [80]:
i. The number of sensors in a WSN can be several orders of magnitude higher than
an ad hoc network.
ii. Sensor nodes are densely deployed in field.
iii. Sensor nodes are expected to failures in WSN.
iv. The topology of a sensor network changes when sensor node failure.
v. WSN mainly use broadcast or tree communication paradigm whereas most ad
hoc networks are based on point-to-point communications.
vi. WSN are limited in power, computational capacities, and memory.
Yick [81] classified the WSN into five types: terrestrial WSN, underground WSN,
underwater WSN, multi-media WSN, and mobile WSN. Terrestrial WSNs typically
consist of hundreds to thousands of inexpensive wireless sensor nodes deployed in a
application area. For deploying WSN, sensors can generally either deterministically or
randomly be placed in an area of interest [82]. The type of sensors determines the
choice of the deployment scheme rely on application and the environment. A viable
controlled node deployment is necessary when sensors are expensive or when their
operation is significantly affected by their position. Such scenarios include
underwater WSN applications, highly precise seismic monitoring, and placing
imaging and video sensors. Moreover, in some applications random distribution of
wireless sensor nodes is the necessary option. This is chiefly fact for harsh
environments such as a battle field or a disaster region. Depending on the node
distribution and the level of redundancy, random node deployment is considered a
good approach that can achieve the required performance goals.
20
3.2 Subsystem of Smart Wireless Sensor Node
In this section, the subsystems of a smart wireless sensor are discussed. Figure
3.1 shows the composition of wireless sensor node. Generally, a smart wireless sensor
is composed of three or four functional subsystems: on-board micro-processor unit,
sensing unit, wireless communication unit, and power unit [12, 83]. The
computational micro-processor was generally used for the computational tasks. For
SHM application, to convert analogue sensor outputs to a digital format, the ADC
with a resolution of 16 bits or higher is needed. The sensing unit always includes an
ADC to connect sensors. For capturing low-order global response modes of a civil
structure, the wireless sensing unit should be designed to record response data at
sample rates at least 100 Hz. To capture high-order response modes of structural
components of civil structure, the wireless sensing unit should therefore need
relatively as high as possible sample rates (greater than 500 Hz) [12, 84]. The random
access memory (RAM) was used to stack the measured and processed data. A flash
memory with software programs was used for the system operation and data
processing [85]. A 16 bit digital sensor data could be stored at one time, at least 256
KB of RAM is required for a WSN application. For the simultaneous multiple tasks
application, at least approximately 256 KB of flash memory would be provided for
that purpose [84]. The wireless communication unit is used to communicate with
other wireless sensor nodes and to transfer the sensing data using a RF radio modem
and antenna Lower data loss rate is necessary for the highly reliable wireless
communication in the conditions of channel interference, multi-path reflections and
path losses. Radio signals always decrease as they propagate through structural
materials such as reinforced concrete or steel material and the large scale civil
structures require wireless communication ranges of at least 100 m. [12, 84, 86].
21
Figure 3.1 Composition of wireless sensor node
3.3 Wireless Sensor Platforms
A number of smart wireless sensor platforms have been developed in academia
and industries. Straser and Kiremidjian [87] first proposed a design of a low-cost
wireless modular monitoring system (WiMMS) for monitoring the civil structures by
integrating a microcontroller and a wireless radio. Lynch et al. [88] used more
powerful computational unit to improve the WiMMS. The WiMMS platform has been
improved further by Wang et al. [89] with a software service which utilize the
computational power of the wireless sensor for multiple tasking (e.g. processing or
transmitting data while collecting data). The tasks can be executed simultaneously.
Aoki et al. [90] have developed remote intelligent monitoring system (RIMS) for
structural health monitoring of bridges and infrastructures. The RIMS consists of
high-clock microcontroller, 3-axis MEMS piezoresistive accelerometer, and
internet-based wireless modem. This system can be controlled via ethernet protocol.
Chung et al. [91] have developed a wireless sensor platform named DuraNode for
22
monitoring bridges and buildings. DuraNode not only has a special feature for
wireless sensing but also allows the wired internet data communication for building
structures using an established Local Area Network (LAN). Farrar and Allen [92]
have developed a smart wireless sensor platform, called Husky. This platform
performs a series of damage detection algorithms by embedding a Java-based damage
detection algorithm package (DIAMOND II). Chen and Liu [93] presented a mobile
agent approach for enhancing the flexibility and reducing raw data transmission in
wireless structural health monitoring sensor networks. In their approach, they
developed an integrated wireless sensor network consisting of a mobile agent-based
network middleware and distributed high computational power sensor nodes. A
mobile agent system called Mobile-C was embedded this mobile agent middleware.
The mobile agent middleware allows a sensor network moving dynamically to the
data source. With mobile agent middleware, a sensor network is able to be
implemented in newly developed diagnosis algorithms and can be adjusted in
response to operational or task changes.
Besides the smart wireless sensor platforms developed in the academia, a number
of commercial smart wireless sensor platforms have been also developed for SHM
applications in the industries. Mica Motes, which is initially developed at the
University of California-Berkeley and subsequently commercialized by Crossbow Inc.,
may be the most famous commercialized wireless sensor platform. The Mote is an
open source wireless sensor platform with both its hardware and software. Motes is
very popular to the public [12]. The Wisden was design for structural data acquisition.
The Wisden perform wavelet-based compression techniques to overcome the
bandwidth limitations based on Mica2 motes platform [94]. Mica mote and
piezoelectric sensors then integrated for a parallel distributed structural health
23
monitoring system. The developed system can successfully monitor the concentrated
load position or a loose bolt position [95]. The performance of the mica mote was
investigated through a two story steel structure testing on shaking table. The results
shows that Mica mote can get the sufficient performance for the intended purpose
[96]. Mote has been successively revised to Imote and Imote2 by Intel. The Imote2
may be the most powerful and promising smart wireless sensor platform for SHM. It
built with 32 bit XScale processor with a RAM of 32 MB and a flash memory of 32
MB, and an integrated radio with a built-in 2.4 GHz antenna. Nagayama et al. [11]
successfully implemented a decentralized computing strategy in a wireless smart
sensor networks that used Imote2 for system identification. In their study, the
measured data was aggregated locally by a selected sensor node within the
communication distance in sensor group, and only limited pre-processed data was sent
back to the base station to locate the health of the structure.
3.4 Factors Influencing Sensor Network Design
Based on lacking of power, having physical damage or environmental
interference, some sensor nodes may fail or be blocked. The failure of sensor nodes
should not affect the overall task in WSN. Therefore, the reliability or fault tolerance
issue should be considered. Robust fault tolerance is the ability to endure sensor
network functionalities without any interruption due to sensor node failures [97].
In studying a different phenomenon, the hundreds or thousands number of sensor
nodes may be deployed. Depending on the application, the number may reach an
extreme value of millions. The new WSN approaches must be able to deal with this
number of nodes. The high density nature of the sensor networks can be less than 10
m in diameter for some specific applications [98].
24
The cost of a single node is very important to influence the overall cost in WSN
with a large number of sensor nodes. If the cost of the WSN is more expensive than
deploying traditional sensors, then the WSN is out of its benefit. It is why the cost of
each sensor node has to be kept low. For example, the novel technology allows a
Bluetooth radio system to be less than 10$ [99]. Also, the cost of a sensor node should
be much less than 1$ in order for the WSN to be feasible in the future.
The size of sensor node may be another important issue in some WSN
applications. The required size may be smaller than even a cubic centimeter which is
light enough to remain suspended in the air [100]. Apart from the size, there are also
some other constraints for designing a sensor node. These nodes must [101]
i. Consuming in extremely low power,
ii. Operating in high volumetric densities,
iii. Having low production cost and be dispensable,
iv. Operating unattended,
v. Be adaptive to the environment.
Sensor nodes are densely deployed to record different phenomenon. Therefore,
they usually work isolated in remote geographic areas. They may be working
i. in busy intersections,
ii. in the interior of a large machinery,
iii. at the bottom of an ocean,
iv. inside a twister,
25
v. on the surface of an ocean during a tornado,
vi. in a biologically or chemically contaminated field,
vii. in a battlefield beyond the enemy lines,
viii. in a home or a large building,
ix. in a large warehouse,
x. attached to animals,
xi. attached to fast moving vehicles, and
xii. in a drain or river moving with current.
3.5 The 802.15.4 and Zigbee Standards
The ZigBee Alliance [102] is an association of companies working together to
develop reliable, cost-effective, low-power wireless networking standard. ZigBee
technology was applied in a wide range of products and applications across consumer,
commercial, industrial and government markets worldwide. ZigBee builds upon the
IEEE 802.15.4 standard which defines the physical and MAC layers for low cost, low
rate wireless application. ZigBee defines the network layer specifications for star, tree
and peer-to-peer network topologies and provides a framework for application
programming in the application layer. Figure3.2 shows the 802.15.4 and Zigbee
Architecture. The following subsections give more details on the IEEE and ZigBee
standards.
26
Figure 3.2 802.15.4 and Zigbee Architecture
IEEE 802.15.4 standard
The IEEE 802.15.4 standard [103] defines the protocol of the physical and MAC
layers for Low-Rate Wireless Personal Area Networks (LR-WPAN). The advantages
of an this standard maintaining a simple and flexible protocol stack that are ease of
installation, reliable data transfer, short-range operation, extremely low cost, and a
reasonable battery life.
The physical layer supports three frequency bands and 27 channels with Direct
Sequence Spread Spectrum (DSSS) access mode: a 2450 MHz band (with 16
channels), a 915 MHz band (with 10 channels) and a 868 MHz band (1 channel), all
using the. The 2450 MHz band employs Offset Quadrature Phase Shift Keying
(O-QPSK) for modulation. The 868/915 MHz bands rely on Binary Phase Shift
Keying (BPSK). In addition to radio on/off operation, the physical layer supports
functionalities for channel selection, link quality estimation, energy detection
IEEE 802.15.4 MAC
Applications
IEEE 802.15.4
2400 MHz
PHY
IEEE 802.15.4
868/915 MHz
PHY
• Network Routing
• Address
translation
• Packet
Segmentation
• Profiles
ZigBee
•Packet generation
•Packet reception
•Data transparency
•Power Management
•Channel acquisition
•address
•Error Correction
27
measurement and clear channel assessment.
The MAC layer defines two types of nodes: Reduced Function Devices (RFDs)
and Full Function Devices (FFDs). The FFDs node are equipped with a full set MAC
layer functions can act as a network coordinator or a network end-device. When
acting as a network coordinator, FFDs using beacons for synchronization,
communication and network join services. RFDs can only act as end-devices which
equipped with sensors/actuators. They may only interact with a single FFD. Two main
types of network topology are considered in IEEE 802.15.4, namely, the star topology
(Fig3.3)and the peer-to-peer topology (Fig3.4).
Figure 3.3. Star topology
Full function device
Reduced function device Communications flow
Master/slave
PAN
Coordinator
Star Topology
28
Figure 3.4. Peer-to-peer topology
The ZigBee standard
ZigBee [102] deals with the higher layers of the protocol stack. The network
layer (NWK) organizes and provides routing over a multihop network (built on top of
the IEEE 802.15.4 functionalities). The Application Layer (APL) provide a interface
for distributed application development and communication. The APL comprises the
Application Framework, the ZigBee Device Objects (ZDO), and the Application Sub
Layer (APS). In the Application Framework, user can defined have up to 240
Application Objects of a ZigBee application. The ZDO provides services that allow
the application objects to discover each other and to organize into a distributed
application. The APS offers an interface to data and security services to all application
objects and ZDO.
ZigBee identifies three device types. A ZigBee end-device is corresponding to an
IEEE RFD or FFD for a simple task. A ZigBee router is an FFD with routing
capabilities. The ZigBee coordinator is an FFD for managing the whole WSN. The
Full function device Communications flow
Point to point Cluster tree
Peer-Peer Topology
29
ZigBee network layer supports more complex topologies like the tree and the mesh
while the IEEE 802.15.4 naturally uses the star topology. Multihop routing, route
discovery and maintenance, security and joining/leaving a network are among the
functionalities provided by the network layer with consequent short (16-bit) address
for assigning to newly joined devices.
3.6 Operating Systems for Wireless Sensor Networks
As the technology moving on, from its root in the Computer Science and
Electronic Engineer community, TinyOS has been adopted as an easy used WSN
operation system in more and more application fields, like Civil Engineering.
TinyOS is designed for low power, ad-hoc, and embedded sensor networks; it is
a component based and event driven operation system which make it a very flexible
and easy to use. These features make TinyOS easy to deploy in different WSN
platforms like Mica, TelosB, Imote2, EYES, BTNode, also Gains and Hawk platform
from China. Although the low level hardware devices are different, researchers who
work on it could use similar environment and share their work results by the
component mechanism.
30
Figure 3.5 Schema of TinyOS (Crossbow)
This is a great advantage for the WSN users. Researcher can use as much as
existed code like the multi-hop algorithm due to the absence of embedded
programming skill. Just as the Figure 3.5 shows, when the researcher wants use the
Multi-hop protocol in his application based on TinyOS, the only thing is to find the
Xmesh Component then adds it to extent the application. Therefore, researchers could
focus on the key problems like data flow, analyses model etc, instead of paying too
much time on the WSN tools itself.
New!
31
Figure 3.6 Architecture of TinyOS ( Crossbow)
The Figure 3.6 shows the architecture of TinyOS. The architecture divide into
three parts, are represented by function A, B and C. The function C means the
hardware device provided by industry, basically consisted with MCU, radio chip,
storage chip, and sensors. The function B means the TinyOS core and Firmware, this
part is also usually provided along with hardware device. The function A is the user
definition part, should be modified by researchers. Also the share of components is
realized in this part.
Hardware Presentation/Abstraction Layer
Mote Devices Sensor Devices
TOS Scheduler (“Main”)
System Components
Library
Components TOS Component Interfaces
Application
Configuration
Application Specific
Components
Application Interfaces
EventsCommands
Hardware Presentation/Abstraction Layer
Mote Devices Sensor Devices
TOS Scheduler (“Main”)
System Components
Library
Components TOS Component Interfaces
Application
Configuration
Application Specific
Components
Application Interfaces
EventsCommands
TOS
User
Legend
HW
A
C
B
32
Figure 3.7 The process in TinyOS( Crossbow)
The process in TinyOS could be generally dived into two categories: event and
task. Event represents a completed sensor reading, or a RF message received, or any
other user specified event. When the event fired, a pre-defined function will be carried
out to process this event. This kind of process should be as quick as possible to avoid
interference among different events. So some long time desired process are classify as
Task will be active when CPU is free, like some data reconstruction and signal
process work should be done in Task.
The regular steps of a TinyOS process are as follows:
i. Physical event fired.
ii. Event handler signaled.
iii. Event parsed and Task posted according to different Event.
iv. Event finished
v. Task active
vi. Carried Task
vii. Task Finished
33
There are two types of component; module and component. Components is
written with code and wired together as a Configuration to create a Mote application.
The interface provides two methods, ―provide‖ and ―use‖, to wire the components.
The table 3.1 shows all description of keywords in TinyOS.
Table 3.1 Description of keywords in TinyOS.
Keword Description
interface A collection of event and command definitions
module A basic component implemented in nesC.
configuration A component made from wiring other components.
implementation Contains code & variables for module or
configuration.
components List of components wired into a configuration.
provides Defines interfaces provided by a component.
uses Defines interfaces used by a module or configuration.
as Alias an interface to another name.
command Direct function call exposed by an interface.
event Callback message exposed by an interface.
Due to the limited memory, the memory usage and memory space need to be
managed. The TinyOS uses the static memory model to manage the memory. The
components of static memory model are statically linked together and size required
determined at compile time. The global variables are used to conserve memory and
the pointers are also using here.
34
TinyOS, the innovation operation system enables researchers to implement their
own WSN research easily, enables researchers share their code and research result all
over the world instead of building everything by oneself, enables an entry level
researcher build up a prototype system very easily and quickly. Plenty of component
based libraries, supporting mesh protocols, distributed services, data acquisition tools
and QoS service, which makes user easy to construct the individual application with
high efficiency and reliability. However, for civil engineers, students and researchers,
using TinyOS to develop the individual SHM application are not very easily. A new
WSN development environment should be developed.
3.7 Time Synchronization
In wireless sensor networks, the data fusion is a basic operation. Data from each
sensor is collected by a data fusion center [104]. The fusion of individual sensor uses
exchanging messages that are time stamped by each sensors local clock. This task
needs an exact time among the sensors. Protocols that provide exact time clock are
synchronization protocols. Over the past few decades, researchers have developed
successful clock synchronization protocols for wired networks. However, these
protocols are unsuitable for a WSN because the challenges posed by WSN are
different and manifold from wired networks.
There is several facts influence synchronization in WSN. The major error in time
synchronization schemes is the message delivery delay with non-determinism in the
latency estimating. A description of sources of synchronization error was first
described by Koeptz [105] and extended recently by Horauer et. al. [106].
The main time synchronization errors are shown below [106]:
(1). Send Time: The delay in the packet traversal from the message assembly
35
at the application layer all the way down to MAC layer. Highly
non-deterministic.
(2). Access Time: Is the channel contention time, that in dense broadcast
medium such as ours can be in the order of hundreds of milliseconds.
Least deterministic part of the message delivery.
(3). Interrupt Handling Time: The delay between the radio chip raising and
the microcontroller responding to an interrupt. Can be an issue if
interrupts are disabled on the microcontroller.
(4). Transmission and Reception Time:The delay in sending or receiving the
entire length of the packet over the channel. Largely deterministic, is a
function of bandwidth and packet size.
(5). Propagation Time: The delay, for a particular symbol of the message, in
traversing all the way to the receiver. The propagation time can be
deterministic if the speed of propagation is assumed constant, and
endpoint location is known, or if synchronization exchange is performed
with assumption of path symmetry. This delay can be significant in the
underwater acoustic channel since assuming clocks will not skew over
packet exchanges would be incorrect.
Basically, there are two schemes to synchronize clocks: Sender-Receiver and
Receiver- Receiver. All synchronization schemes developed within these two basic
frameworks. Network Time Protocol (NTP) is widely used in the Internet. It is
approach works well with high latency and high variability [107]. The NTP protocol
estimates both offset and skew using long-term, bi-directional exchange of time
information. Unfortunately, NTP is a poor match for WSN for several reasons. First, it
36
assumes communications are relatively inexpensive. However theWSN are bandwidth
and energy constrained. Next, NTP is designed for constant low rates operation in the
background. (At a maximum polling rate of 16 sec, NTP took around an hour to
reduce error to about 70us). On the other hand, TSHL (Time Synchronization for High
Latency)[108] exchanges number of broadcast beacons to compute skew and then
perform one bidirectional exchange to compute a skew-corrected offset. Generally,
TSHL and NTP synchronize clock using the same time information, however TSHL
reduces energy consumption using a smaller number of broadcast beacons by
replacing long-term bidirectional communication.
37
CHAPHER 4 INTEGRATED WSN-based SHM
SYSTEM
4.1 Introduction
Characterized by its low manufacturing cost, low power requirement,
miniaturized size, and no need for cabling, the wireless sensor networks (WSN) has
emerged as an attractive sensing technology for deploying dense distributed
sensors[12, 13]. Among the various wireless monitoring techniques [12-17], the
MICA mote is a commercially available product that has been used extensively by
researchers and developers. It utilizes an Atmel ATmega 128L processor that runs at 4
MHz. The 128L is an 8-bit microcontroller that has 128 kilobytes of onboard flash
memory to store the mote's program. MICA motes have certain limitations, including
limited sampling rate, processing ability, storage, and transmission ability. In
comparison, the advanced wireless sensor platform Imote2 is considered as a better
choice for developing and deploying customized wireless sensor networks efficiently
[109, 110]. While the WSN-based sensing system is accessible and achievable,
development beyond a prototype towards full-scale structure applications often incurs
hardware and software complexities, making it quite challenging, especially for civil
engineers with limited WSN experience. Also, robust system development is
problematic under a complex programming environment. To this end, this thesis
presents an easy-to-use development environment based on the .NET Micro
Framework (NETMF). Applications can be implemented efficiently under the
NETMF. Evidently, the accelerated development of the SHM techniques can be partly
attributed to the user-friendly environment brought by the WSN.
38
This chapter discusses the architecture for proposed Integrated WSN-based SHM
system. This system is developed based on NETMF platform, which is an advanced
sensor platform compatible with NETMF. Architecture, wireless sensor node design,
energy harvesting system, software framework, and WSN-Based SHM service to be
employed in this thesis are described.
4.2 Architecture of Integrated WSN SHM System
The proposed integrated WSN-based SHM system is shown in Fig.4.1. This
system is based on hybrid wireless communication, structural conditions will be
assessed remotely. The proposed architecture consists of sensing nodes, cluster head
nodes, transfer node, and base station. The purpose of sensing node is to measure the
responses of structure or the environmental parameters. They are deployed in
structure to perform specific tasks such as sensing, data processing, and
acknowledgement. Each sensing node is based on Imote2 platform which is
comprised of a microprocessor, sensor module, and communication device. All
sensing data will be sent to a cluster head nodes in their own community using one
hop transmission.
Cluster head node can communicate with the sensing nodes or other cluster
heads of the neighboring communities to exchange information or triggering sensing
task. The cluster head is a dule core design which combine Imote2 platform and
second embedded device. The cluster head has extra wireless module and GPS. The
extra wireless module has more powerful RF power can increase the wireless
communication distance. The GPS is useful for synchronization and localization
purposes. The transfer node is functions as coordinate node for managing the cluster
heads and hoping the data. The sensing data will be sent to the base station, if needed,
39
using fast communication device such as WiFi. The base station is the highest level
end device that has largest memory, most powerful processor and highest
communication capability. The base station node is the gateway between smart sensor
networks and the Host computer. The base station uploads the data to remote
monitoring and control server via satellite, 3G telecommunication or WiMAX
communication. A Host computer may be needed in this architecture as an interface to
sends commands and parameters to smart sensor networks via the base station.
Figure 4.1 Integrated WSN-based SHM system architecture: different roles are
assigned to nodes
4.3 Hardware Design
This section shows all hardware design of wireless sensor nodes which are
proposed and employed in this study. The proposed wireless sensor nodes are sorted
into sensing nodes, cluster head nodes, transfer node, and base station.
40
4.3.1 Sensing Node
The sensing node consists of mote component and sensor component, is shown
in Fig. 4.2. Figure 4.3 indicates the block diagram of sensing node architecture. The
―mote‖ usually contains Microprocessor, Memory storage, and RF chip. The Imote2
platform is chosen for mote component herein. The Imote2 platform is an advanced
platform for wireless sensor network nodes. The Imote2 platform is suitable for
application in a structural health monitoring system that typically requires signal
pre-processing functions applied to sensing nodes [83].
Figure 4.2 Photo of sensing node
The Imote2 platform contains the PXA271 (XScale) processor. This processor
operates in the low-voltage (0.85V) and low-frequency (13 MHz) mode, thereby
41
enabling low-power operation. The processor can be scaled to 104 MHz at the lowest
voltage and increased to 416MHz using dynamic voltage scaling. The Imote2
platform can perform complex signal processing functions, such as FFT and image
compressing. The Imote2 platform contains 256 KB of SRAM divided into four equal
64KB banks. The PXA271 of the Imote2 platform is a multi-chip module that has
three chips in a single package—the processor, 32 MB of SDRAM, and 32 MB of
flash memory. Such a memory improvement is a significant advantage compared with
the 4kB of RAM in Mica motes. The sensing task must store a massive amount of
sample data in external flash memory. The write/read data rate is limited by one slow
hardware device. However, the Imote2 platform stores sample data in RAM instead of
the complex and slow flash write function. The processor integrates many I/O options,
making it extremely flexible in supporting different sensors, A/Ds, and radio options.
These I/O options are I2C, three synchronous serial ports, three high-speed UARTs,
GPIOs, SDIO, USB client and host, AC97 and I2S audio codec interfaces, a fast
infrared port, PWM, a camera interface, and high-speed bus. The Imote2 platform
uses the CC2420 IEEE 802.15.4 radio transceiver, which supports a 250 Kbps data
rate with 16 channels in the 2.4GHz band.
42
Figure 4.3 Block diagram of sensing node architecture.
Table 4.1 compares mote platforms which are widely used in many research
groups. It can be seen in the table that the transmission range of RIRS is 300 m
outdoor, two to three times than other motes. The power consumption of TelosB is
much lower than other motes. For SHM application, high sampling rate sensing
always get great amount of data. By comparison, although Imote2 does not
outperform other motes in power consumption and transmission range, the larger
memory is more suitable for SHM application. The larger memory is beneficial for
smart sensor applications, which is require multiple tasks, such as wireless
communication, data logging, data acquisition, and in-situ signal processing.
The sensor component is ITS400 herein. The ITS400 sensor board is designed to
connect to the basic connectors on the Imote2. It contains a ST Micro LIS3L02DQ 3d
accelerometer, SHT15 temperature/humidity sensor, TAOS TSL2651Light Sensor and
43
Maxim MAX1363 4 Channel General Purpose A/D. This Accelerometer board has a
range of +/- 2g with 12 bit resolution. It communicates to Imote2 with two possible
interfaces, SPI or I2C. By default, the sensor is connected to SSP1 on the Imote2.
On board SHT15 sensor can be used for requiring high accuracy temp reading
(+/- 0.3 degC) and humidity. This sensor interfaces to the Imote2 through two GPIO
pins. The data pin of the SHT11 is connected to GPIO 100, whereas the clock pin is
connected to GPIO 98. The TAOS TSL2651 light sensor interfaces to the Imote2
through the I2C bus. The interrupt pin is connected to GPIO99 through a NAND gate.
If GPIO 99 conflicts with another board, the BT_TXD pin can be used instead by
loading R35. The board includes a Maxim MAX1363, 4 channel, 12 bit resolution
general purpose ADC for quick prototyping. Each channel supports 0-3 V input
signals. The ADC interfaces to the Intel Mote 2 through the I2C bus. The analog pins
are brought out to a Molex PN-39357-0003 connector (J5).
Table 4.1 Comparison of mote platform
MICA2 MICAZ IRIS TelosB Imote2
Processor ATmega128L ATmega128L ATmega128L MSP430 XScalePXA271
Clock
speed(MHz) 7.37 7.34 16 6.7 13-416
Bus size(bits) 8 8 8 16 32
Program Flash
Memory(bytes) 128 k 128 k 128 k 48 k 32 M
EEPROM (bytes) 4 k 4 k 4 k 16 k
RAM (bytes) 4 k 4 k 4 k 10 k 256 k+ 32 M
external
ADC 10 bit 10 bit 10 bit 12 bit N/A
Active power(mA) 8 8 8 1.8 31@13 MHz
44
Sleep power(uA) 15 15 8 5.1 390
Measurement
Flash(bytes) 512 k 512 k 512 k 1024 k 32 M
RF Chip CC1000 CC2420 At86rf230 CC2420 CC2420
Frequency(MHz) 2400-2483.5 2400-2483.5 2400-2483.5 2400-2483.5 2400-2483.5
Data rate (kbps) 38.4 250 250 250 250
Transmission
power (dbm) -20-5 -24-0 3 -24-0 -24-0
Outdoor range 150 m 100 300 100 100
Indoor range 30 m 30 50 30 30
Typica supply
voltage 3.3 V 3.3 V 3.3 V 3.3 V 4.5 V
Size (mm) 58 x 32 x 7 58 x 32 x 7 58 x 32 x
7
65 x 31 x
6 36x 48 x 9
Weight (g) 18 18 18 23 12
4.3.2 Cluster Head and Transfer Node
The cluster head node is a dule core design which combines Imote2 platform and
second embedded device FEZ domino platform, which contains extra wireless module
and GPS. Figure 4.4 and 4.5 show the Block diagram of Cluster Head node
architecture and photograph of Cluster Head node on a box. Cluster head nodes are in
charge of collecting data from sensing nodes and managing data; these nodes have
45
features such data storage on SD card or USB memory devices, database support,
Runtime Loadable Procedure (RLP), which is used to implement intensive data
processing function and time-critical routines such as FFT. The Transfer node has the
same hardware design to cluster head node without Imote2.
Domino is a tiny open-source board based on ARM7 microcontroller from NXP
running Microsoft NETMF. With this sophisticated combination, a developer can
easily control this microcontroller IOs and interfaces such as SPI, UART (Serial Port)
and I2C with simple unified managed code (C# code). Many libraries have already
included similar FAT file system, threading, USB Client, USB Host, UART, SPI, I2C,
GPIO, PWM, ADC, DAC, and many more.
The extra wireless module used herein is Digi XBee modules. XBee is a wireless
communication module that Digi built to the 802.15.4/ZigBee standard. These
modules communicate with the microcontroller using the UART. Table 4.2 gives the
specification of Xbee models. It can be seen in the table that the transmission range
can be designated by choosing specific model. It is useful for deploying WSN in large
scaled structure.
46
Figure 4.4 Block diagram of Cluster Head node architecture.
Table 4.2 Specification of Xbee models.
Model XB XB-Pro XB-ZB XB-ZB-Pro XB-868 XB-900 XB-XSC
Protocol 802.15.
4 802.15.4 Zigbee-Pro Zigbee-Pro RF RF RF
Frequenc
y 2.4GHz 2.4GHz 2.4GHz 2.4GHz
868MH
z
900MH
z 900MHz
Tx power 1mW 100mW 2mW 50mW 315mW 50mW 100mW
Sensitivity -92dB -100dBm -96dBm -102dBm -112dB
m
-100dB
m -106dBm
Range 500m 7000m 500m 7000m 40Km 10Km 24Km
47
Figure 4.5 Photograph of Cluster Head node on box
4.3.3 Base Station
The base station is the highest level device that can store a mass of data and
upload to remote center. The base station is based on ChipworksX Module.
ChipworkX Module is a small (67.6mmx47mm) ARM9 processor based board with
SO-DIMM200 slot interface and a very high performance. ChipworkX Module
supports a complete set of features such as FAT, USB device and many more. In
addition, it supports many other exclusive features, for example, USB host, PPP,
GPRS, 3G...etc. The sensing data can store and access using SD/MMC cards and USB
memory devices such as Thumb Drives and Hard Disks. Moreover, the USB host
function can access other devices like mice, keyboards, joysticks, printers, USB
modem and more. Furthermore, using SQLite database, allowing fast logging and
retrieving of standard SQL quires. For real-time and high processing needs, Runtime
48
Loadable Procedures, allow users to load their own compiled native code (C or
assembly).
Figure 4.6 Block diagram of base station architecture.
4.3.4 Multi-Sensor Board Design
This study developed a new sensor board which integrates a six degrees of
freedom inertial measurement unit (6 DoF IMU) SD746 sensor chip (SensorDynamics
Inc.) and a GPS module. The SD746 is used to measure acceleration and the GPS is
used to synchronize the sensing nodes. Figure 4.7 shows the simplified block
diagram of proposed 6 DoF-GPS sensor board. The chosen 6 DoF IMU is a
temperature-compensated and calibrated sensor that integrates functionality of 3-axis
gyro sensor and 3-axis accelerometer in one small package. Exposing the module to
accelerations and angular rates leads to small variations of the MEMS sensors
49
capacitances. Amplifier and ADC convert these changes in the gyro and accelerometer
capacitances into digital signals and then processed by dedicated Digital Signal
Processing (DSP) units. The DSP applies filtering, bias and sensitivity adjustment,
temperature compensation. Individual sensor configuration data are stored into a
non-volatile memory which contains also the individual serial number for production
and traceability. SPI and I2C communication interfaces share the same digital IOs. A
set of internal monitoring signals, continuously available though both SPI and I2C
interface allows the user to check the correct functionality of the entire module.
Figure 4.8 presents the photograph of the developed multi-sensor board. Table 4.3 and
4.4 shows the physical and electrical characteristics of gyro and accelerometer sensor.
Figure 4.7 Simplified Block Diagram of Multi-sensor board
u-bloxneo-5Q
GPS
3-Axis Gyro Sensor
3-Axis Accelerometer
ADC
UART
DSP SPI
UEXT connector
GPS bee
SD746
50
Figure 4.8 Photography of developed Multi-sensor board
Table 4.3 Physical and Electrical Characteristics of Gyro Sensor
PARAMETER UNIT MIN TYP MAX
Full Scale Range °/s - ±20481 -
Resolution °/s/LSB - 0.0621 -
Digital Resolution bit - 16 -
Noise Density °/s/√Hz - 0.06 -
Nominal Bandwidth (programmable) Hz 10 - 80
Bandwidth Accuracy % -10 - 10
Bias at Room Temperature °/s - ±5 -
Bias Variation Over Temperature2 °/s - ±5 -
Sensitivity Error at Room Temp. % - ±2 -
Sensitivity Error Over Temperature3 % - ±5 -
Linearity Error % - ±0.2 -
Cross Axis Sensitivity1 % - ±2 -
Signal Update Rate KHz 10 - -
51
Table 4.4 Physical and Electrical Characteristics of Accelerometer Sensor
PARAMETER UNIT MIN TYP MAX
Full Scale Range g - ±84 -
Resolution μg/LSB - 244.1 -
Digital Resolution bit - 16 -
Noise Density (mg)/√Hz - 0.28 -
Nominal Bandwidth (programmable) Hz 10 - 80
Bandwidth Accuracy % -10 - 10
Bias at Room Temperature g - ±0.05 -
Bias Variation Over Temperature2 g - ±0.2 -
Sensitivity Error at Room Temp. % - ±2 -
Sensitivity Error Over Temperature3 % - ±5 -
Linearity Error % - ±0.5 -
Cross Axis Sensitivity % - ±2 -
Signal Update Rate KHz 10 - -
The GPS module is designed based on NEO-5Q GPS chip (Ubox Inc.) and used
to synchronize the sensing nodes. The accuracy of timepulse signal is 40ns RMS. The
NEO-5Q has a 32-channel acquisition engine with over 1 million effective correlators
and is capable of massive parallel searches across the time/frequency space. This
enables a Time To First Fix (TTFF) of less than 1 second while long correlation/dwell
times make possible the best-in-class acquisition and tracking sensitivity. An available
functionality is KickStart, a new feature enabling accelerated acquisition of weak
signals. Once acquired, satellites are passed on to a power-optimized dedicated
tracking engine. This arrangement allows the GPS engine to simultaneously track up
to 16 satellites while searching for new ones. It has several features that are shown
below:
i. 50-channel u-blox 5 engine with over 1 million effective correlators
ii. <1 second Time To First Fix for Hot and Aided Starts
iii. -160dBm acquisition and tracking sensitivity
iv. Accelerated startup at weak signals for modules with KickStart feature
52
v. Supports AssistNow Online and AssistNow Offline A-GPS services; OMA
SUPL compliant
vi. 4 Hz position update rate
vii. Miniature 2.0mm pitch header, compatible with Xbee socket
4.3.5 Power Module
Power sources and power consumption are the critical issue in WSN if batteries
have to be periodically replaced. Hence, this study proposed a power module with
energy harvesting system and energy storage unit. The proposed theory is harvesting
the ambient energy like the vibration of the bridge to power the sensor device or to
recharge the primary battery. A novel energy harvesting based on piezoelectric
transducer was developed. Next section completed describes the proposed energy
harvesting system in detail.
4.4 Windmill-Magnet Integrated Piezoelectric (WMIP) Energy
Harvesting System
Although a wireless sensor network based structural health monitoring system
has many benefits, power sources and power consumption are the critical issue if
batteries have to be periodically replaced. Therefore, harvesting and storing the
ambient sources of energy to supply the sensor node seems to be an agreeable
approach. The sources of ambient energies typically include thermal, sunlight, wind,
RF and vibration energy. This study only considered the vibration based piezoelectric
energy harvesting approach. Several researches proposing and reviewing the possible
energy harvesting schemes can be found in the literature [111-114].
53
4.4.1 General Theory of Vibration-Based Energy Harvesting
Method
A single degree of freedom lumped spring mass system is mostly utilized to
model a vibration-based energy harvesting system. A diagram of a piezoelectric
cantilever beam with a proof mass at the end and a model of equivalent lumped spring
mass system is shown in Fig.4.8. The equation of motion of a single degree of
freedom (SDOF) system consisting of a mass m, a spring with spring constant k, and a
damping coefficient c under external excitation is described as
( ) ( ) ( ) ( )mz t cz t kz t my t (4.1)
where z is the net displacement. Assuming that the external excitation is harmonic
given as ( ) siny t Y t , the steady –state solution for the mass displacement is
given by
2
22 2
( ) sin
1 2
n
n n
z t Y t
(4.2)
where the is the phase angle given by
1
2tan
c
k m
(4.3)
The approximate mechanical power of a piezoelectric generator can be given by
54
3
2 3
22 2
1 2
n
n n
m Y
P
(4.4)
The maximum energy can be extracted by setting the excitation frequency to
match the natural frequency of the system and is given by
2 3
4
nmYP
(4.5)
Observing Eq. (4.5), it provides evidence that power can be optimized by
lowering damping, increasing natural frequency, mass, and amplitude of excitation.
Figure 4.9 (a) Cantilever beam with tip mass, (b) model of equivalent lumped
spring mass system
4.4.2 Energy Harvesting Circuits
A typical energy harvesting circuit, with piezoelectric generator, battery and
55
sensor, is shown in Fig.4.10. The piezoelectric material produces an ac voltage output
when the piezoelectric deformed. Therefore, this voltage needs to be converted to a dc
voltage before charging the capacitor. The four diodes form a bridge circuit to
perform a rectifier. Energy harvesters typically produce small amounts of energy over
long periods, therefore an energy storage component in the form of a supercapacitor
usually contain in harvesters system. A larger capacitor provides power for a longer
time for the same load but takes more time to charge it. Typical supercapacitor often
has a much lower voltage than standard electrolytic capacitors, a zener diode usually
used to prevent the voltage across the supercapacitor from increasing beyond its
maximum voltage rating. When applying a load, the supercapacitor discharging
immediately, and the voltage across the supercapacitor starts dropping. Hence, a
dc/dc-voltage-converter IC is used to assurance a fixed voltage at the output.
Figure 4.10 Typical piezoelectric harvesting circuit
The typical piezoelectric harvesting model can be improved via new energy
harvesting IC, LTC3588-1 (LINEAR TECHNOLOGY). This IC combines a low-loss
Piezoelectric generator
rectifier
dc/dc-voltage-
converterSupercapacitor
Sensor Battery
56
full-wave bridge rectifier with a high efficiency buck converter to form a tiny energy
harvesting component for an ambient energy harvesting. The peak load currents are
much higher than a piezoelectric generator can produce. Thus the LTC3588-1
accumulates energy that released to the load to supply the sensor or controller. For the
different purpose of applications, this IC provides four output voltages, 1.8V, 2.5V,
3.3V and 3.6V via two selectable pins. Typically, up to 100mA of continuous output
current can be provided; however, a higher output current can be supplied by way of
sizing the output capacitor. Base on this advanced energy harvesting IC, the
architecture of energy harvesting based wireless sensor node was shown in Fig.4.11. A
LTC3588-1 connects with piezoelectric generator and backup battery to supply the
sensor. This architecture proposed that when ambient energy is obtainable, the battery
is unloaded, but when the ambient source ceases, the battery initiates and serves as the
backup power supply. The backup battery must connect with a series blocking diode
connected to VIN pin to prevent reverse current flowing into the battery. Any stack of
batteries can be used as long as the battery voltage does not exceed 18V. One should
be considered that the peak voltage of piezoelectric generator should exceed the
battery voltage. This approach not only decreases the replacing time of battery, but it
also improves reliability and elasticity of sensing system. For instance, an energy
harvesting sensor node is deployed on a structure such as a bridge, may gather energy
when the vehicles pass through the bridge. However, in off-peak times the vehicles
are rarely and vibration is also low, a battery backup can still supplies the sensor node.
57
Figure 4.11 Architecture of energy harvesting based wireless sensor node
4.4.3 Windmill-Magnet Integrated Piezoelectric Energy
Harvesting System
Most researchers tested piezoelectric energy harvesting device on machine or
motor. The reason is the regular high frequency vibration of structure such as machine
or motor can give maximum efficiency of energy harvesting. However, the low
frequency vibration of civil structure such as bridge limits the energy harvesting
efficiency. Therefore, this study proposed a scheme to solve this barrier. In first stage,
the idea is to transfer the rotation of windmill to vertical vibrate the piezoelectric
beam for energy harvesting. The proposed simple concept is shown in Fig. 4.12 Three
piezoelectric materials were tested for the proposed concept shown in table 4.5.
Battery
Piezoelectric generator
Wireless SensorNode
58
Table 4.5 Different types of piezoelectric product
Type Dimension(length X
width)
Manufacturing
Company
PZT 60X20mm APC
MFC 85X28mm Smart Material
Raw Volture 100X25.4mm MIDE
Figure 4.12 Schematic diagram of windmill based piezoelectric harvesting system
Fig. 4.13 shows the experimental setup on overpass in Tokyo University. A
piezoelectric film was set as a cantilever beam mounted on a base which was fixed in
the bridge. Wind can drive the windmill to beat the piezoelectric film to output the
voltage. The Fig. 4.14 to 4.16 illustrates the output voltage of PZT, MFC and Raw
Volture respectively. As shown in Figure, the peak voltage represents the high voltage
output when the windmill beat the beam. The result also shows the MFC has high
voltage under the same excitation. The result confirms that the proposed windmill
based piezoelectric energy harvesting concept is practical. However, contact between
windmill and piezoelectric beam might stop the whirling of blade. To solve this
problem, a novel windmill-magnet integrated piezoelectric (WMIP) energy harvesting
system was proposed in Fig. 4.17. The proposed WMIP harvesting system integrates a
windmill, magnet and piezoelectric beam for energy harvesting. A magnet was
Transfer Mechanism
Piezoelectric beam
Transfer the rotation to vertical vibrate the Piezoelectric
59
bonded on the tip of piezoelectric beam as a mass. A windmill was placed a proper
distance from the piezoelectric beam and each blade was attached to a magnet. As the
windmill turned, the magnetic force between the piezoelectric beam and windmill
caused the beam to vibrate. Figure 4.18 is an example shows that the proposed WMIP
energy harvesting system gives high frequency (close to 50Hz) and regular output
voltage. This high frequency and regular output voltage provide efficient energy
harvesting and storage.
Figure 4.13 Experiment setup of windmill based piezoelectric harvesting system:
(a) Raw Volture, (b)MFC, (c)PZT, (d) Overpass in Tokyo University.
61
Figure 4.16 Output voltage of Raw Volture
Figure 4.17 Windmill-magnet integrated piezoelectric (WMIP) energy harvesting
system
62
Figure 4.18 Output voltage and FFT of WMIP energy harvesting system
4.5 Three-Tier Software Framework
The proposed software system of integrated WSN-based SHM system is a
three-tier-framework, i.e., node, logging and processing tiers. Herein, the node tier
consists of developed application functions and is installed on all sensor nodes. The
63
application functions are developed in a Microsoft NETMF environment. The
NETMF and Crossbow’s API, i.e., application interface provided by manufacturing
company of Imote2, assist most libraries in dealing with all peripherals and drivers,
allowing us to develop individual applications for the SHM efficiently. The logging
tier, implemented based on C# and NETMF, is installed on the cluster head node and
base station and is intended for logging the data and data storage. The processing tier
is developed using LabVIEW. This tier has several functions. First, raw data are
converted into engineering units and then stored in a database. Next, a customized
user interface is performed to select the desired sensor node from the data. The
sensing data can also be analyzed by an advanced signal processing tool, e.g., filtering,
smoothing, denoise, FFT, wavelet, and further data analysis.
Figure 4.19 The proposed software architecture
64
4.5.1 Net Micro Frameworks-Based Embedded Development
Environment
There are five layers in the basic architecture of NETMF. The platform
adaptation layer (PAL) and hardware abstraction layer (HAL) layer deal with the
hardware. The PAL supports higher-level integration, including asynchronous
communication calls, high-level timers, and list and data structures. The HAL
provides generic access to all peripherals and drivers. The NETMF’s common
language runtime (CLR) layer serves the purpose to load and execute managed code.
By supporting managed code, the NETMF CLR allows for rapid development and
safe execution of application code using modern programming languages and tools.
Microsoft developed several basic application libraries in library layer, and then
researchers can use it directly. As aforementioned description, the NET Micro
Framework help for dealing all peripherals and drivers, the developer hence can focus
on their individual application.
Figure 4.20 shows the structure of Class library which is branch out into three
basic class of NETMF, basic class library of Imote2, and SHM class. The SHM class,
a specialized class, is developed by this study for SHM application. Table4.6 gives a
description of these three classes. Based on these three classes, a SHM application for
specific purpose can be easily developed. Listing 4.1 presents an example code for
periodical deep sleep and sampling. For example, the radio can be initialized via the
constructor in Radio class: public Radio (ushort freq, ushort power, ushort
pan_address, ushort address) where define the frequency, power, node address,
and PAN address. After the parameters, such as the sampling rate, data types, and data
length are declared, the sampling starts remaining 60 s with a sampling rate of 200 Hz.
65
The sampling data then writing to a packet. After sampling, the packet is sent using
the Send method. Furthermore, for power saving, the mote can be put in deep sleep
mode (drawing 525 uA) for 60 s just call the DeepSleep method. Therefore, by using
above main functions and other fundamental APIs, the application can be developed
rapidly.
Figure 4.20 Structure of Class library
66
Table 4.6 Description of Class library
Namespace of class Name of class Description of class
Basic class of NETMF
Mscorlib A subset of the core .NET
classes.
System Only the System.Net
namespace
Microsoft.SPOT.Native Core .NET Micro
Framework classes.
Microsoft.SPOT.Hardware Managed hardware
drivers.
Microsoft.SPOT.Graphics Low-level graphics
classes.
Microsoft.SPOT.Net Internal socket drivers
basic class library of
Imote2
Platform For low-level access to the
mote hardware
Radio To set various radio
options
Sensors Various sensor drivers
Utilities Contains a number of
utility classes
SHM class
Extended math class Extend the basic math
class
Sampling Sampling the sensor data
Synchronization Synchronizing the nodes
FFT To calculate FFT
FRF To calculate FRF
67
Listing 4. 1. Sample code for periodical deep sleep and sampling
public class XAccel
{
public static void Main()
{
//Configurations
const ushort _rfChannel = (ushort)RadioChannel.Ch15; // channel 11
const ushort _rfPower = (ushort)RadioPower.M0DBM; // -10dbm
const int packet_length = 64;
const int packet2_length = 30;
Radio radio = new Radio(_rfChannel, _rfPower, 0xFFFF, 0xFFFF);
byte[] packet = new byte[packet_length];
byte[][] packetall = new byte[50000][];//
short accelX, accelY, accelZ; //accel raw values
int samplingrate = 5; //
samplepakage = 1200;// pakegets
int sendperiod = 50;
//Sampling
for (; ; )
{
while (pt < samplepakage)
{
accelX = (short)(accelX + _sensor.AccelXRaw);
accelY = (short)(accelY + _sensor.AccelYRaw);
accelZ = (short)(accelZ + _sensor.AccelZRaw);
packetall[row][index++] = (byte)(Av_x1 & 0xFF);
packetall[row][index++] = (byte)((Av_x1 >> 8) & 0xFF);
packetall[row][index++] = (byte)(Av_y1 & 0xFF);
packetall[row][index++] = (byte)((Av_y1 >> 8) & 0xFF);
packetall[row][index++] = (byte)(Av_z1 & 0xFF);
packetall[row][index++] = (byte)((Av_z1 >> 8) & 0xFF);
}
}
// sending data then go to deep sleep
radio.SetRadioOption(RadioOption.LocalAddress, (ushort)_nodeid);
for (row = 0; row < samplepakage; row++)
{
for (index = 4; index < packet_length; index++)
{
packet[index] = packetall[row][index];
}
packet[1] = (byte)row;
System.Threading.Thread.Sleep(sendperiod);
radio.Send((ushort)0xFFFF, (ushort)0xFFFF, packet);
}
pm.DeepSleep(600);
}
} //main
}
68
4.5.2 Data Logging and Data Storage
The logging tier, implemented based on C#, is installed on the cluster head node
and base station and is intended for logging the data and data storage. The data is
extracted from each packet and written to a text file or store in a database. The packet
carries three parts, which are TosMsg header, sensor board header and data payload
(Fig.4.21). The TosMsg header carries 5-byte TinyOS compatible header that provides
Imote2 protocol compatibility with MICA and IRIS motes. Table 4.7 Shows the
TosMsg header content. The next group in packet is the sensor board header that
provides basic information of node. Next are the data payload that includes all the
information serial number, timestamp, and sensor data. Herein, sensor data denotes 3
readings from X/Y/Z axis of the accelerometer sensor, each is 2 bytes. There are 10
groups in one packet. Table 4.8 and 4.9 present the content of sensor board header and
data payload.
Figure 4.21. Packet format
Table 4.7 TosMsg header contents
Type Field Name Description
uint16_t Address Destination address of the packet
uint8_t AM type TinyOS Active Message ID
uint8_t Group ID TinyOS group ID
uint8_t Length Length of the packet
TosMsg Header Sensor board header Data payload
Addres AM type Group ID Length
Serial number Timestamp Sensor data
Board_id Packet_id Node_id
TosMsg Header
Sensor board header
Data payload
69
Table 4.8 Sensor board header content
Type Field
Name Description
uint8_t board_id XSensor packet type field (Sensor streams are always a
constant value of 0x11)
uint8_t packet_id XSensor packet sub-type field (XAccel is always a constant
value of 0x01)
uint16_t node_id Network short address of sending node.
Table 4.9 Data payload
Type Field Name Description
uint16_t serial number Incrementing sequence number for yield calculation
uint16_t timestamp Timestamp information
uint16_t accel_x Accelerometer reading of X axis
uint16_t accel_y Accelerometer reading of Y axis
uint16_t accel_z Accelerometer reading of Z axis
…
uint16_t accel_x Accelerometer reading of X axis
uint16_t accel_y Accelerometer reading of Y axis
uint16_t accel_z Accelerometer reading of Z axis
4.5.3 User Interface and Visualized Signal Processing
The processing tier is developed using LabVIEW. This tier has several functions.
First, raw data are converted into engineering units and then stored in a database. Next,
a customized user interface is performed to select the desired sensor node from the
data. The sensing data can also be analyzed by an advanced signal processing tool,
e.g., filtering, smoothing, denoise, FFT, wavelet, and further data analysis. Figure
70
4.22 shows the functions of proposed user interface. Figure 4.23 shows how the
sensor data can be represented in a Figure. Chart and nodes distribution can also be
conveniently observed. Figure 4.25 presents the damage location in a visualization
window which can indicate the location of damage visibly.
Figure 4.22 Functions of proposed user interface
71
Figure 4.23 Sensor data representation and node distribution
Figure 4.24 Signal processing interface
72
Figure 4.25 Damage location representation
4.6 Reliable Data-Sensing and Transmission Service
In this section, WSN-based middleware services for SHM application are
developed. The middleware services include reliable data-sensing and transmission,
which are data aggregation, reliable communication, and synchronized sensing. These
middleware services are implemented on the nodes running .NETMF and are
generally applicable to SHM application.
The reliable data-sensing and transmission service is presented in Figure 4.26.
First, after the sensing nodes and cluster head node are initialized, the cluster head
node sends an inquiring packet to confirm whether the sensing nodes are ready. The
sensing nodes and cluster head node then exchange timestamp packets to synchronize
together based on a two-phase synchronization scheme [108]. In general, two main
clock errors must be corrected, i.e., skew and offset. In the first phase, the proposed
protocol models the skew of all sensing node’s clock; each node is then skew
73
synchronized. Next, the skew is estimated by performing linear regression over
multiple timestamp packets from the cluster head node. Each timestamp packet Pi
contains the transmit timestamp ,B it obtained at the MAC level, just before the
packet leaves the cluster head node. The sensing nodes receive the packet at absolute
time ,B i B St D where B SD refers to the unknown propagation delay between the
cluster head node and the sensing nodes. The sensing nodes then record their local
time ,( ).S B i B SLT t D Although the local time includes the error due to clock skew
and offset in addition to propagation delay, the skew of the local clock can still be
modeled with respect to the reference clock of the cluster head node by performing
linear regression on the difference between ,B it and ,( ).S B i B SLT t D For N packets,
the clock skew can be modeled by linear regression over the data set of (xi,yi) pairs,
where xi and yi are , ,( ( ))B i S B i B St LT t D and ,( )S B i B SLT t D , respectively.
In the second phase, the clock offset is corrected by the classical two-way
synchronization exchange in a manner similar to that of the TPSN synchronization
protocol [115]. Once a sufficient number of packets to estimate the skew are obtained,
the sensing nodes send a packet with the skew-corrected local timestamp,
1 ( 1),ST LT T to the cluster head node. The cluster head node records its received
local time 2 ( 1 )B S BT LT T D , then replies with a packet to sensing nodes with T2
and transmit timestamp T3. Upon receiving the packet, the sensing nodes record the
skew-corrected local time 4 ( 3 ).S B ST LT T D Finally, the sensing node can
compute its clock offset as ( 2 1) ( 4 3) / 2Offset T T T T .
Following completion of the synchronization, all sensing nodes starts to sample
data and write data to an array. The parameters, such as the sampling rate, data type,
74
and data length, are declared before the sampling starts. Following the sampling, the
time-scheduling data transmission procedure is initiated to send data from the sensing
nodes to the cluster head node. Initially, the sensing nodes wait until a sending-delay
time is equivalent to (total sampling time + total packet sending time)* (node ID). For
instance, if the node ID is 0, the sensing node 0 transmits the packet immediately
since the sending-delay is zero. The data in each sensing node are then taken from the
array, filled in a packet, and sent to the cluster head node. With this procedure, each
node sequentially sends data to the cluster head node with a respective sending-delay
to avoid packet collision. However, packets losses occur occasionally, even though the
time-scheduling data transmission is designed to avoid packets collision. The base
station thus continuously checks the number of packets in the sequence from the
sensing nodes and records the number of each missing packet in the sequence. As a
result, the base station can request a sensor node to rectify the missed packets.
75
Figure 4.26. Reliable data-sensing and transmission service
Initializing
· Waiting for receiving packet form base station
Sensing node Cluster head node
Synchronization
· Skew estimation
· Compute its clock offset
· Correct clock
Inquiry
· Inquire of sensing nodes whether they are ready
Acknowledge
· Return acknowledge packet
Synchronization
· Exchange timestamp packet with sensing node
Parameters Declaration(E.g. sampling rate, data
type, data length, etc.)
Sampling dada
Write data to array
End of sampling
Yes
Take out data form array
Fill in packet
Send packet
End of Sending
Yes
Sampling
Time-scheduling data transmission
Waite sending-delay time
Sending-delay=(total sampling time + total
packet sending time)* (node ID)
No
No
Send missing data· Send missing packet to cluster head node
Receive data
· Receive and save data from sensing node
· Record the missing packets
Require missing data
· Send Request Packet for missing data
· Receive and save data from sensing node
Initializing
· Ready for sending commands
76
CHAPHER 5 DEVELOPMENT OF
GLOBAL-LOCAL-INTEGRATED DAMAGE
DETECTION APPROACH
5.1 Introduction
Structural health monitoring (SHM) was broadly classified into local and global
categories. The Non Destructive Testing (NDT) technique is mostly used in local
damage detection. There are many methods for NDE such as ultrasonic, infrared
thermography, eddy current X–ray. Such methods can detect tiny or incipient damage;
however, it is usually expensive, power hungry and bulky. Global SHM, on the other
hand, can discover damage large enough to influence the properties of the entire
structure or large sections of it. Most existing global damage detection methods
identified damage based on changes in basic modal properties that extract from
structural response by low cost sensing systems.
Both local and global SHM each has its own benefits and shortcoming, hence
integrating aforesaid two methods can obtain more effective damage detection result
than only using one of them. This research proposed a global-local-integrated damage
detection approach which is shown in Fig. 5.1.
77
Figure 5.1 global-local-integrated damage detection approach
In the first stage, the global method is adopted to detect the change of
characteristic of entire structure. Subsequently, roughly damage areas of structure are
Structure
Excitations
Sensing System
Response
Signal
processing
Global damage
detection
Local damage
identified
Roughly damage
Yes
No
Final estimation
• Ambient• Earthquake• Shaker• etc.
Building and Civil infrastructure
Integrated WSN SHM System
Substructure-based FRFapproaches
Electro-Mechanical-Impedance (EMI)-Based
damage detection method
78
identified. In stage two, after recognized the damage occurring and approximate
location, a local method was developed to identify the detailed condition of damage.
For instance, in a building structure, a global SHM can roughly indicate the location
of damaged floor. The local SHM then check the component of building structure in
detail. Substructure-based frequency response function approach was proposed as
global damage detection approach. Electro-Mechanical-Impedance (EMI)-Based
damage detection method is used to identify local damage of structure in this study.
5.2 Substructure-Based Frequency Response Function for Global
Damage Detection
A frequency response function (FRF) expresses the structural response to an
applied force as a function of frequency. This response may be represented in terms of
displacement, velocity, or acceleration [116, 117]. Theoretically, FRF can be
expressed in terms of system properties of mass, stiffness, damping, and modal
properties. Moreover, using measured FRF-data has a major benefit that the measured
FRF-data can provide much more damage information in a desired broadband
frequency range than modal data because the modal data are identified mainly from a
very limited number of FRF-data around resonance frequency [118, 119].Accordingly,
an FRF scheme is reasonably expected to be feasible for detecting structural damage.
Several studies have applied the FRF to locate damage. Thyagarajan et al. [120]
developed a method based on FRF data and an optimal number of sensors on a
structure to identify damage locations, overcoming the former limitation of large
number of computations required for high dof. structure. Lee and Shin[121] combined
an FRF-based structural damage identification method (SDIM) with a reduced domain
79
approach to detect damage to beam structures. This method depends on data for the
natural frequencies and mode shapes of the intact beams and the FRFs of damaged
beams to identify structural damage. They demonstrated the feasibility of this SDIM
using numerically simulated damage-identification tests. Their analytical results
illustrated that the use of FRFs seems to be very promising for structural damage
identification. Other investigations have extended FRF methods, such as the FRF
curvature method or the FRF shape-based method, to improve the detection of
damage locations. Sampaio et al.[122] developed a theoretical FRF curvature method
and evaluated the efficiency of this method using numerically simulated data and
experimental data for a real bridge. Their analytical results demonstrated that their
FRF curvature method was effective in detecting damage on single site. Maia et
al.[123] also presented an FRF curvature-based damage detection method and
compared its performance with that of a conventional mode shape-based method.
According to their results, methods based on FRF curvatures outperformed mode
shape-based methods. Liu et al.[124] developed an FRF shape-based method. This
method utilized the imaginary parts of FRF shapes of a beam structure to identify the
damage location before and after damage. Their analytical results exhibited that this
method was effective for a low-frequency range.
As well beam structures, FRF have been applied to detect damage in a building
structure. By using measured FRF and neural networks (NNs), Ni et al.[125]
identified the seismic damage of a 38-storey building model. That study determined
FRFs for dimensionality reduction and noise elimination based on principal
component analysis (PCA). According to their results, PCA is feasible for filtering
unwanted measurement noise. Their damage identified results obtained after moderate
earthquakes or super-strong earthquakes, in which considerable damage as identified
80
at the bottom portion of the building structure. Furukawa et al.[126] developed a
damage detection method using uncertain FRFs based on a statistical bootstrap
method and then applied it to a building structure. Their analytical results confirmed
that the proposed statistical method to detect structural damage can determine whether
the detected damage is a real damage or an identification error due to measurement
noise. Kanwar et al.[127] demonstrated the feasibility of using FRF to the structural
damage of reinforced concrete buildings. By using FRF, Hsu and Loh[128] detected
damage of building structure subjected to earthquake ground excitation. This method
required FRFs of intact and damaged systems as well as system matrices of the intact
system to derive the damage identification equations. Their numerical results
indicated that only the frequencies close to the natural frequencies of the damaged
system need to be selected in order to solve the identification equations. Meanwhile,
their experimental studies demonstrated the feasibility of the proposed algorithm in
detecting the structural damage within an acceptable accuracy.
Accordingly, an FRF scheme is reasonably expected to be feasible for detecting
structural damage. An novel substructure-based FRF approach with a damage location
index (SubFRFDI) was proposed to localize damage to building structures from
seismic response data.
5.2.1 Substructure-Based Frequency Response Function Approach
A frequency response function (FRF) is a transfer function, expressed in the
frequency domain. FRF can be represented in terms of magnitude and phase with real
and imaginary components. The response may be given in terms of displacement,
velocity, or acceleration. Consider a linear system as represented by the diagram in
Figure 5.2.
81
Figure 5.2 Frequency Response Function Model
The relationship in Figure 5.2 can be represented by the following equations
( )( )
( )
XH
F w
(5.1)
where the ( )F is the input force as a function of the angular frequency . ( )H is
the transfer function and ( )X is the displacement response function. Each function
is a complex spectral function, and considers each to be a Fourier transform for
simplicity.
Consider a SDOF system which was subjected to a force excitation and the basic
model of an SDOF system consists of a mass m, a spring with spring constant k, and a
damping coefficient c. The equation of motion is
( ) ( ) ( ) ( )mx t cx t kx t f t (5.2)
( )( ) ( ) ( )
c k f tx t x t x t
m m m (5.3)
By convention
22 ,n n
c k
m m (5.4)
where n is the natural frequency in (radians/sec), and is the damping ratio.
Transfer function
( )H
Input Force Displacement Response
( )F ( )X
82
By substituting Eq.(5.4) in to Eq. (5.3),
22 ( )
( ) 2 ( ) ( ) nn n
f tx t x t x t
k
(5.5)
Takes the Fourier transform of each side of equation (5.5), the resulting FRF is
2
2 2
( )
( ) ( (2 ))
n
n n
X
F w k j
(5.6)
where the n , k and are the nature frequency, stiffness and damping ration,
respectively. The defined FRF uses displacement as the response. It is known as
receptance.
By replacing the displacement response ( )X with velocity ( )X and
acceleration ( )X , two different types of FRFs cane be defined as:
2
2 2
( )
( ) ( (2 ))
n
n n
jX
F w k j
(5.7)
2 2
2 2
( )
( ) ( (2 ))
n
n n
X
F w k j
(5.8)
where the Eqs.5.7 and 5.8 are Mobility FRF and Accelerance FRF, respectively.
The reciprocals of the three FRFs of an SDOF system also have useful physical
significance and can be represented as
83
( )Dynamic stiffness =
( )
force F w
displacement X (5.9)
( )Mechanical impedance =
( )
force F w
velocity X (5.10)
( )Apparent mass =
( )
force F w
acceleration X (5.11)
In a shear building, beams and floor systems are assumed to be rigid in flexure.
Several factors, such as the axial deformation of beams and columns, and the effect of
axial force on column stiffness, are neglected in analysis. Although an ideal shear
building does not exist in practice, it is helpful for illustrating the use of
substructure-based FRF to locate damage. For a damped multi-degree of freedom
(MDOF) shear building with, say, N dofs., the equations of motion are
( ) ( ) ( ) ( )M x t C x t K x t f t
(5.12)
where M , C , and K are N N mass, viscous damping, and stiffness
matrices, respectively; and ( )x t and ( )f t are 1N vectors of the displacement
functions and effective external excitation load, respectively. The mass matrix is
diagonal, while C and K are tri-diagonal. Applying the Fourier transform to Eq.
(1) and performing a simple arrangement yields
1
2X K i C M F
(5.13)
which can be written simply as
84
( )
XH
F (5.14)
where ( )H is an FRF of the MDOF structure system as similar to the FRF
of SDOF system.
The MDOF structure is divided herein into various substructures. Consider an
N-storey shear building subjected to base excitations with acceleration ga , which can
be described by a simplified model (Fig. 5.3). N substructures of the original structure
can be easily established. The first substructure has the 1st—N
th dofs. while the second
has the 2nd—N
th dofs. Accordingly, the i
th substructure has the i
th—Nth
dofs. (Fig. 5.3).
The responses of any degrees of freedom of the ith
substructure are the same as those
of the original complete structure.
When 1i , considering an undamped sheer building model. For the ith
substructure, the balance of forces in the ith
floor can be written as
1 1 1( ) ( ) ( ( ) ( )) ( ) 0
i i i i i i i ik x t x t k x t x t m x t
(5.15)
where ( )ix t is the total displacement response of the ith
dof., and
( ) ( )i gx t x t a .
For the (i+1)th
to Nth
floors, the balance of forces is
1 1 2 2 1 1 1( ) ( ) ( ( ) ( )) ( ) 0
i i i i i i i ik x t x t k x t x t m x t
(5.16)
2 2 1 3 3 2 2 2( ) ( ) ( ( ) ( )) ( ) 0
i i i i i i i ik x t x t k x t x t m x t
(5.17)
85
1( ) ( ) ( ) 0
N N N N Nk x t x t m x t
(5.18)
The Eqs. (5.15’)~(5.18’) can be written as
1 1 1 1( ) ( ) ( ) ( ) ( )
i i i i i i i i ik k x t k x t m x t k x t
(5.19)
1 1 2 1 2 2 1 1( ) ( ) ( ) ( ) 0
i i i i i i i i ik x t k k x t k x m x t
(5.20)
2 1 2 3 2 3 3 2 2( ) ( ) ( ) ( ) 0
i i i i i i i i ik x t k k x t k x m x t
(5.21)
1( ) ( ) ( ) 0
N N N N N Nk x t k x t m x t
(5.22)
Equations (5.19’)~(5.22’) can be written as a matrix equation.
1 1
0 ( )
( )
0 ( )
i i
i i
N N
m x t
m x t
m x t
1 1 1
11 1 2 2
( ) ( )
0 ( ) 0
( )0 0
ii i i i i
ii i i i
N N
x tk k k k x t
x tk k k k
k x t
(5.23)
When damping is considered, Eq.(5.23’) can be expressed as
1 1
0 ( )
( )
0 ( )
i i
i i
N N
m x t
m x t
m x t
(5.24)
86
1 1
1 1 2 2 1
( )
0 ( )
0 ( )
i i i i
i i i i i
N N
c c c x t
c c c c x t
c x t
1 1 1 1
11 1 2 2
( ) ( ) ( )
0 ( ) 0
( )0 0
ii i i i i i i
ii i i i
N N
x tk k k c x t k x t
x tk k k k
k x t
where ( )ix t is the total displacement response of the ith
dof., and ( ) ( )i gx t x t a .
87
Figure 5.3. (a) Original complete structure, (b) the ith
substructure.
Equation (5.24) represents a single input/multiple output system. Clearly, the
substructure-based FRFs of the jth
dof. in the ith
substructure are
( )
1
( )( )
ji
j
i i i
XH
i c k X
, j = i to N (5.25)
where jX and 1iX are the Fourier transforms of ( )jx t and1( )ix t
, respectively. The
substructure-based FRF given by Eq. (5.25) depends on the properties of the ith—N
th
dofs. To estimate( ) ( )i
jH from Eq. (5.25), ic and ik must be known. Unfortunately,
(b)
xi-1(t)
mi
mi+1
mN
xi(t)
xi+1(t)
xN(t)
ki, ci
ki+1,ci+1
kN, cN
m1
mi-1
x1(t)
k1,c1
Feff(t)
xi-1(t)
mi
mi+1
mN
xi(t)
xi+1(t)
xN(t)
ki, ci
ki+1,ci+1
kN, cN
ith substructure
(a)
Original complete structure
ki-1, ci-1
88
ic and ik are usually unknown in the damage diagnosis process. Moreover,
acceleration responses are normally measured in monitoring the responses of a
structure in an earthquake. Therefore, the measured acceleration responses of the
original complete structure are used herein and these substructure-based FRFs are
further simplified as
( )
1
( )ji
j
i
XH
X
, j = i to N (5.26)
where jX and 1iX are the Fourier transforms of ( )jx t and 1( )ix t , respectively. A
specified substructure-based FRF,( ) ( )i
jH , can be estimated from measured
acceleration of the (i-1)th
and jth
dofs. Notably, when i =1, referring to the original
complete structural system, these substructure-based FRFs are given by
(1) ( )j
j
g
XH
A , j = 1 to N (5.27)
where gA is the Fourier transform of ga .
Theoretically, when the damage is assumed to have occurred in the column(s)
between the ith
and (i-1)th
dofs. (such that ki is reduced and ci is increased), the
substructure-based FRF is significantly altered in the ith
dof., as described by,
( ) ( )i
iH . Likewise, if stiffness ki and kl decline simultaneously, then the change in the
corresponding FRFs, ( ) ( )i
iH and( ) ( )l
lH , is greater than what would typically be
observed in ( ) ( )i
jH . Consequently, damage can be identified as having occurred at
89
a single or multiple sites. For efficiency, only one substructure-based FRF, ( ) ( )i
iH ,
is determined herein for each substructure to reduce the computational time. Damage
of the shearing building structures is located based on the FRFs,(1)
1 ( )H ,
(2)
2H , …, ( )N
NH , of all substructures.
Based on the aforementioned description, change in the substructure-based FRF
is related to damage and can be utilized as an essential index to locate damage of the
building structures. The operating conditions of this work are based on a known initial
state, continuously measured seismic response data, and comparisons of
before-and-after damage scenario for a shear building. Initially, sensors are deployed
on the shear building and the corresponding response data are measured immediately
after an earthquake excitation. The initial stiffness of the structure is then defined and
the corresponding FRFs
(1)
1, ( )uH ,(2)
2, ( )uH …,
( )
, ( )N
N uH are computed as a known
undamaged state, also referred to as a before-scenario state. Next, the sensors
continuously collect seismic response data of the structure after each subsequent
earthquake and the particular FRFs (1)
1, ( )dH ,(2)
2, ( )dH …,
( )
, ( )N
N dH are obtained
following, referred to as an after-scenario state. Before and after damage scenarios are
compared to determine the damage location since a change in the substructure-based
FRF is related to damage. A dissimilarity between the substructure-based FRFs in
damaged and undamaged states can be used to identify the damage. The absolute
dissimilarity ( )iP is defined as
90
( ) ( )
, ,( ) ( ) ( )i i
i i d i uP H H (5.28)
where ( )
, ( )i
i dH and ( )
, ( )i
i uH are the magnitudes of ( )i
iH in the damaged and
undamaged states, respectively. These N dissimilarities 1( )P — ( )NP can be
correspondingly calculated for a shear building with N floors. To locate conveniently
and quantify the damage, a substructure-based FRF damage location index
(SubFRFDI) for the ith
substructure is proposed. It is given by
221 ( ) /b
i i
a
SubFRFDI Exp NDF n
(5.29)
where , a, b, and n are working parameters and the ( )iNDF is expressed as
( ) ( )
, ,max max
( )( )
max ( ) , ( )
ii
i i
i d i u
PNDF
H H
(5.30)
The coefficient is a control value that scales the index between zero and one
and is set to five in this work. The range of selected frequencies for calculating
SubFRFDI is set to a—b, where a is a starting frequency of zero and b is the end
frequency, which equals the first modal frequency (undamaged state). The aforesaid a
and b are determined by trial-and-error method. The value n equals (b - a) divided by
sampling time, which is 0.005 s in this work. Figure 5.4 shows the FRF of
substructure.
91
Figure 5.4. FRF of Substructure
If the properties of a structural system do not change, then the iSubFRFDI is
close to zero. However, if the damage to storey i in a shear building is severe, then the
value of iSubFRFDI is high. The resulting apparently significant peak value(s) of
SubFRFDI indicate(s) that damage occurs at either a single site or at multiple sites.
For instance, if damage occurs only in the ith
dof., then only the SubFRFDIi from all
SubFRFDI has a significant peak value. If damage occurs at more than one site, such
as in the ith
, jth
, and kth
dofs., then the corresponding SubFRFDIs will all have
significant peak values.
5.2.2 Numerical Study
This section presents a numerical example of a six-storey shear plane frame
structure that is subjected to base excitation, to evaluate the feasibility of the proposed
damage-detection scheme. The structural parameters of each floor were the same -
Damaged
Undamaged
a b
92
mass mi = 30 kg and stiffness ki = 55.5 kN/m (i = 1—6). Numerical simulation was
performed with 5% modal damping. The damage in this work was simulated as
reduced storey stiffness. Single-site and multiple-site damage were studied. Table 5.1
presents the simulated damage cases: Dam_ki (i = 1—6) denotes damage in the form
of reduced stiffness ki at a single site i; similarly, Dam_ki&kj (i ≠ j) represents
damages in the form of reduced stiffness ki and kj at multiple sites. Table 5.1 also
presents the percentage reduction in ki in parentheses.
Table 5.1 Descriptions of simulated damage cases.
Damage
case 1 2 3 4 5 6 7 8
Notation Dam_k1 Dam_k2 Dam_k3 Dam_k4 Dam_k1 Dam_k3 Dam_k3
&k5
Dam_k1
&k5
Damage
level k1 (15%) k2 (15%) k3 (15%) k4 (15%)
K1
(15%-30
%)
k3
(15%-30
%)
K3 (15%)
and
K5 (15%)
k1 (15%)
and k5
(15%)
Damage
case 9 10 11 12 13 14 15 16
Notation Dam_k1
&k3
Dam_k1
&k2
Dam_k1
&k2
Dam_k1
&k3
Dam_k1
&k3
Dam_k1
&k3&k5
Dam_k1
&k3&k5
Dam_k1
&k3&k5
Damage
level
k1 (15%)
and k3
(15%)
k1 (10%)
and
k2 (20%)
k1 (10%)
and
k2 (20%)
k1 (20%)
and
k3 (10%)
k1 (10%)
and
k3 (20%)
k1 (15%)
k3 (10%)
and k5
(5%)
k1 (5%)
k3 (15%)
and k5
(10%)
k1 (10%)
k3 (15%)
and k5 (5%)
For damage case 1, Figure 5.5 compares the conventional FRFs of the structure
in the undamaged and damaged states, while Fig. 5.6 compares the substructure-based
FRFs for each substructure. These Fig.s present only the parts of an FRF that
correspond to frequencies that are less than the second resonant frequency of the
complete structure or each substructure. The reduction in stiffness in the first storey
93
markedly changed the conventional FRFs of each floor whereas significant changes
were observed only in the substructure-based FRF of substructure 1. The
substructure-based FRF characterized and located damage more effectively than the
conventional FRF.
To apply the proposed substructure-based FRF approach to locate structural
damage, the locations of damage were determined using the SubFRFDI. Figure
5.7-5.10 presents the identified damage locations in four single-damage cases, 1, 2, 3,
and 4. When damage occurred only to the ith
storey, the value of the SubFRFDI that
corresponds to the ith
substructure greatly exceeded those that correspond to the other
substructures. The proposed index is verified to have located the damaged storey.
Four different percentage reductions in stiffness were assessed for damage case 1 and
3. Figure 5.11 and 5.12 shows the identified damage locations and corresponding to
damage extent. As shown in Fig., the damage can be correctly located and the value
of SubFRFDI was increasing corresponding to damage extent.
95
Figure 5.5 Comparisons of conventional FRFs of the structure in undamaged
and damaged states for Case 1.
98
Figure 5.7 The identified damage locations in single-damage cases 1.
Figure 5.8 The identified damage locations in single-damage cases 2.
99
Figure 5.9 The identified damage locations in single-damage cases 3.
Figure 5.10 The identified damage locations in single-damage cases 4.
100
Figure 5.11 The identified damage locations in single-damage cases 1 for
different damage extent.
Figure 5.12 The identified damage locations in single-damage cases 3 for
different damage extent.
To study the effects of the frequency range in the SubFRFDIi, Fig. 5.13 presents
101
the values of the SubFRFDI of each substructure in damage case 3, obtained using
two different frequency ranges. The frequency ranges, a—b, were set to 0—1.6 Hz
and 0—2 Hz. The former range covers frequencies lower than the first modal
frequency, while the latter range includes frequencies above the first modal frequency.
The maximum value of the SubFRFDI for various substructures was normalized to
one for comparison. The value of the SubFRFDI for substructure 3 was markedly
larger than the others, when the values of SubFRFDI were computed using the two
frequency ranges (Figure 5.13), indicating that the third storey was damaged.
However, when the values of the SubFRFDI were obtained using the frequency range
0—2 Hz, the damage on the second storey may be mistakenly identified.
Consequently, the damaged storey is easily identified when the SubFRFDI value is
estimated using the substructure-based FRFs using frequency ranges below the first
modal frequency. Those results further demonstrate that change-related information in
the first model is adequate to identify the alteration between undamaged and damaged
states, explaining why use of the first intact modal frequency in this work is
acceptable.
To reflect the fact that measured data always contain noise, 5% and 10% white
noise were respectively added to the numerically simulated responses and input
acceleration in damage case 3. Figure 5.14 presents the effects of noise on
substructure-based FRF. Figure 5.15 presents the effects of noise on SubFRFDIi
estimations. Although noise influences SubFRFDIi values, especially for large i, the
SubFRFDIi was still, correctly, greatest at i = 3.
To demonstrate that the proposed index is independent to responses to various
input earthquakes and only reliant to the structural properties, the Kobe and El Centro
102
earthquakes were applied as input excitations. Figure 5.16 presents the SubFRFDI
values of different substructures in damage case 4. The SubFRFDI values varied only
slightly with the base excitations.
To compare the proposed SubFRFDI with an existing FRF-based damage
location identification method, Fig. 5.17 presents the analytical results for damage
case 3 obtained using the proposed method and the FRF curvature-method-based
index (FRFCDI)[122]. The FRFCDI method is an extension of the mode shape
curvature method for locating damage. The FRFCDI is given by
( ) ( )i i
i U DFRFCDI
(5.31)
where ( )i
U and ( )i
D are the FRF curvatures of the ith
dof. of an undamaged
structure and the corresponding damaged structure, respectively. Since FRFCDIi
and SubFRFDIi have different values in a particular damage scenario, the maximum
values of FRFCDIi and SubFRFDIi were normalized to one for comparison. The
proposed SubFRFDI outperforms the FRFCDI method (Figure 5.17). The FRFCDI
does not locate damage correctly, but the SubFRFDI does. Additionally, for the
highest storey, FRFCDI has a larger error than the SubFRFDI.
103
Figure 5.13 The values of the SubFRFDI of each substructure in damage case 3,
obtained using two different frequency ranges.
Figure 5.14. The FRF of substructure: With and without noise.
Number of Substructure
1 2 3 4 5 6
Norm
ali
ze S
ub
FR
FD
I
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Chosen Range of
Frequency:0-1.6Hz
Chosen Range of
Frequency:0-2Hz
104
Figure 5.15 The effects of noise on SubFRFDIi estimations for Case 3.
Number of Substructure
1 2 3 4 5 6
Su
bF
RF
DI
0.00
0.02
0.04
0.06
0.08
Excitation_Kobe
Excitation_EL Centro
Dam_k4
Figure 5.16 The SubFRFDI values of different substructures in damage case 4
with different earthquake excitations.
Number of Substructure
1 2 3 4 5 6
Su
bF
RF
DI
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Without Noise
With 5% Noise
With 10% Noise
Dam_k3
105
Number of Substructure
1 2 3 4 5 6
No
rma
lize
Su
bF
RF
DI
0.0
0.2
0.4
0.6
0.8
1.0
1.2
SubFRFDI
FRFCDI
Dam_k3
Figure 5.17 The analytical results for damage case 3 obtained using the proposed
method (SubFRFDI) and the FRF curvature-method-based index (FRFCDI).
The proposed approach can identify damage locations at multiple sites and a
single site. Figure 5.18-5.27 presents the SubFRFDI values for substructures in
damage cases 7—16. The corresponding peaks reveal that even though multiple
damage sites suffered damage of various degrees. Figures 5.18-5.27 present the
proposed index accurately determined the locations of the damage in a multi-damage
scenario.
106
Figure 5.18 The SubFRFDI values for substructures in damage cases 7, multiple
damage.
Figure 5.19 The SubFRFDI values for substructures in damage cases 8, multiple
damage.
107
Figure 5.20. The SubFRFDI values for substructures in damage cases 9, multiple
damage.
Figure 5.21 The SubFRFDI values for substructures in damage cases 10, multiple
damage.
108
Figure 5.22 The SubFRFDI values for substructures in damage cases 11, multiple
damage.
Figure 5.23 The SubFRFDI values for substructures in damage cases 12, multiple
damage.
109
Figure 5.24 The SubFRFDI values for substructures in damage cases 13, multiple
damage.
Figure 5.25 The SubFRFDI values for substructures in damage cases 14, multiple
damage with various damage level.
110
Figure 5.26 The SubFRFDI values for substructures in damage cases 15, multiple
damage with various damage level.
Figure 5.27 The SubFRFDI values for substructures in damage cases 16, multiple
damage with various damage level.
111
5.2.3 Implement of SubFRFDI in Integrated WSN-Based SHM
System
An SubFRFDI-based protocol was developed and embedded in the integrated
WSN-based SHM System. Figure 5.28 presents the SubFRFDI-based protocol when
using the WSN-based SHM System. Initially, the sensing nodes and base station
exchange timestamp packets so that they can be synchronized using a two-phase
synchronization scheme. The radio can be initialized via the constructor in Radio
class of C# where define the frequency, power, node address, and PAN address.
During the first phase, the proposed protocol models the skew in the clocks for all
sensing nodes; each node is then skew synchronized. Next, the skew is estimated by
performing linear regression over multiple timestamp packets from the base station.
After synchronization, the sensing node waits for the fire command from event node.
When the earthquake occurs, the event node immediately broadcasts a packet with a
fire command to the sensing nodes. To send and receive the packet is easily using the
classes Send and ReceiveAny. Finally, the sampling procedure for all sensing nodes
starts performing sampling data and writing data to a response array. Corresponding
parameters such as sampling rate, data type, and data length, are declared before
sampling starts.
After sampling, the FRF calculation procedure is initiated. The sensing node i-1
then send FFT data to sensing node i. The FFT data are written to an input array. The
FRF is next estimated by a FRF calculation function and is locally stored. Initially, the
FRF is estimated as a known undamaged state. After each subsequent earthquake, the
FRFs are referred to as a damage state. After the undamaged and damaged FRF are
obtained, the SubFRFDI can then be estimated in all sensing nodes. Finally, all
112
sensing nodes send their SubFRFDI to the base station to evaluate the damage to the
structure. In this protocol, the base station node and event node are powered by wall
plug to be always waking. The sensing nodes are powered by rechargeable battery
with energy harvesting system. All sensing nodes are set to be deep sleep mode and
waking up by turns. At least one sensing node is waking on each floor for ensuring
measuring. The mote can be put in deep sleep mode (drawing 525 uA) just call the
DeepSleep class for a specified period of time.
113
Figure 5.28 An SubFRFDI-based protocol
Initializing· Waiting for fire command from event node
Sensing node i
Parameters Declaration(E.g. sampling
rate, data type, data length, etc.)
Sampling dada
Write data to Response array
End of sampling
Yes
Sampling
No
FRF calculation
Estimate FFT using response array data
Receive FFT data from node i-1
Write data to input array
Estimate FRF=response /input
Calculate damage index
First time to
calculate FRF
Set FRF= ( )i
uH
Set FRF= ( )i
dH
Event trigger· Broadcast fire command to sensing nodes
Event node
Send index
· Send SubFRFDI to base station
Sensing node i-1
YES
NO
Synchronization· Skew estimation
· Compute its clock offset
· Correct clock
Synchronization
· Exchange timestamp packet with sensing node
Receive damage index
· Receive SubFRFDI from all sensing nodes
Base station
Control flow
Wireless communication between
nodes
221 ( ) /b
i i
a
SubFRFDI Exp NDF n
114
5.3 Electro-Mechanical-Impedance (EMI)-Based Local Damage
Detection Method
5.3.1 Fundamental of EMI-Based Model
The EMI based damage detection approach use the EM coupling effect between
active surface-bonded Piezoelectric patches and host structural. Researchers [64, 71]
showed that the EMI is related to the mechanical impedance of a host structure, thus
allowing monitoring the properties of the host structure using the measured electrical
impedance. A piezoelectric patch-structure bonded system can be modeled as a circuit
system [129], shown in Fig. 5.29. Herein, piezoelectric patch is modeled as a
capacitor (Cp) and a self sensing-actuation voltage source (Vp) caused by input
voltage (Vin). The output voltage, couple with Vp and Vin, can be expressed as
( )( ) ( ( ) ( ))
( ) ( )
Rout in p
R p
ZV V V
Z Z
(5.32)
where ZR is the electrical impedance of the resister and Zp is electrical impedance of
piezoelectric patch. Subsequently, the electrical impedance of piezoelectric patch can
be written as
( ) ( )( ) ( ) 1
( )
in p
p R
out
V VZ Z
V
(5.33)
115
Figure 5.29. Diagram of PZT-structure bonded system
Since the self sensing-actuation voltage source (Vp) is related to structural mass,
as confirmed by numerous researchers. Eq. 5.33 indicates that the Zp has significant
response to structural damage. A piezoelectric patch-structure bonded system can be
further modeled as an electro mechanical admittance (EMA) model (Inverse of EMI).
The electro mechanical admittance model can be expressed as [130]
2 2
33 31 31
tan( )ˆ ˆ( )pmT E E
pm s
Zwl klY j d Y d Y
t Z Z kl
(5.34)
where w, l, and t are the width, length, and thickness of the piezoelectric patch; ˆ EY
is the complex Young’s modulus of the piezoelectric patch at zero electric field; 33
T
is the complex dielectric constant of piezoelectric patch; 2
31d is the coupling
piezoelectric constant in the x direction at zero stress; ˆ/ Ek Y is the wave
structure
PZT
Vin VoutR
VpCp
AC
V
12
3
w
l
t
Piezoelectric patch
116
number that is related to mass density , ˆ EY ; and excitation frequency ,
respectively; pmZ and sZ are the mechanical impedances of the piezoelectric patch
and the host structure, respectively. For a SDOF structure system, the mechanical
impedances of structure is defined as
0 i
s s
FZ Z e
(5.35)
in which
2
2
2s
m kZ c
(5.36)
where the 0F is a harmonic excitation force at angular frequency ; is velocity
response in frequency domain; the m, c and k are mass, damping and stiffness of
structure, respectively.
Equation (5.34) indicates that the EMI (or EMA) is directly related to the
mechanical impedance of structure. As described in Eq. (5.34) and Eq. (5.35), when
damage occurred to structure, the mechanical impedance of structure will be changed.
Therefore, if the mechanical impedances of the piezoelectric patch remain undamaged,
any changes in the EMI signal correlates with the damage in structure.
Although the EMI provides a qualitative method for detecting structural damage,
the quantification approach need be established. A simple statistical algorithm, which
is based on frequency-by-frequency comparisons, referred to Root Mean Square
Deviation (RMSD), was used to develop the quantitative assessment of damage in
previous research [69, 129, 131]. The RMSD is defined as
117
2
2100
bD U
i i
i a
bU
i
i a
I I
RMSD
I
(5.37)
where U
iI and D
iI are the real part of impedance of piezoelectric patch at the ith
frequency point in undamaged and damaged structures, respectively. In a RMSD
damage metric chart, the greater numerical value of the metric, the larger the
difference between the baseline reading and the subsequent reading indicates the
presence of damage in a structure.
5.3.2 Measurement of Piezoelectric-Impedance
The measurement of piezoelectric impedance is typically implemented by an
impedance analyzer. The disadvantages of using impedance analyzer are relatively
bulky size, expensive, and no portable. The Peairs et al. [132] developed an effective
device using a digital signal analyzer with an FFT function that can measure and
record the electric impedance of a PZT. Advantages of the new impedance-measuring
device include its low cost, greater accessibility, and smaller size.
Credit to the progress of Micro Electronic Mechanical System (MEMS), the new
single-chip impedance measurement device, AD5933 has been developed by Analog
Device Inc. The AD5933 is a high precision impedance measurement device that
combines an on-board frequency generator with a 12-bit, 1 MSPS, analog-to-digital
converter (ADC). The frequency generator allows exciting external complex
impedance with a known frequency. The response signal from the impedance is
sampled by the on-board ADC and a discrete Fourier transform (DFT) is processed by
an on-board DSP engine. The DFT algorithm returns a real (R) and imaginary (I)
118
data-word at each output frequency. Once calibrated, the magnitude of the impedance
and relative phase of the impedance at each frequency point along the sweep is easily
calculated. This is done off chip using the real and imaginary register contents, which
can be read from the serial I2C interface. The block overview is shown in Fig. 5.30.
Figure 5.30. Block overview of AD5933
5.3.3 Selection of Frequency Ranges
The sensitivity of damage detection using NDE techniques is related to the
frequency band selected. Initial damage in structure does not result in any measurable
change in the structure’s global stiffness properties. Hence, high frequency excitation
is typically used. It was found that a frequency range with a high mode density covers
more structural dynamic information. The range for a given structure typically used in
the impedance methods is in the range of 30 kHz to 250 kHz and determined by a trial
and error method. In the impedance-based method, numbers of peaks are usually
chosen since there is a greater dynamic interaction over that frequency range. A higher
frequency range (higher than 150 kHz) is found to be favorable in localizing the
119
sensing, while a lower frequency range (lower than 70 kHz) covers more sensing
areas. [71, 131, 133].
Under the high frequency ranges used in this impedance-based method, the
sensing region of the piezoelectric patch is localized to a region close to the
sensor/actuator. Through various case studies in Esteban’s work [134], it has been
estimated that (depending on the material and density of the structure) the sensing
area of a single PZT can vary anywhere from 0.4 m (sensing radius) on composite
reinforced concrete structures, to 2 m on simple metal beams. The frequency ranges
higher than 500 kHz have been found to be unfavorable, because the sensing region
becomes extremely small and the PZT sensors show adverse sensitivity to their
bonding conditions or PZT itself rather than behavior of a structure monitored.
120
CHAPHER 6 EXPERIMENTAL VERIFICATION
6.1 Introduction
In this chapter, this dissertation focused on the experimental study. When the
theoretical basis of damage detection method and wireless SHM system based
measuring system are completed, series experimental study are conducted to verify
the proposed algorithm and system. First off all, the integrated WSN-based SHM
system is verified involving sensor calibration, wireless communication quality, data
losing, interference in material, power consumption. Subsequently, the proposed
damage detection approach with proposed SHM system then investigated in a ¼ -scale
six-storey steel structure that was designed by the National Center for Research on
Earthquake Engineering (NCREE), Taiwan. Subsequently, the proposed wireless
SHM system is employed in a bridge structure to test the feasibility in field.
6.2 Verification of Integrated WSN-Based SHM System
6.2.1 Verification of Sensing Node
In this test, the reliable data-sensing and transmission service is implemented on
the sensing node and cluster head node. The sensing capability is calibrated compared
to reference wire sensor. Two sensing nodes and a reference wire sensor (CXL01LF1)
are attached on a Quanser shaking table to measure acceleration of table. The shaking
table is excited by controller using Sin wave excitation input. Two sensing node and
reference sensor are installed next to each other and the corresponding signals from
those are expected to be identical. Measured dynamic responses were acquired by
121
sensing nodes and sent to the cluster head node connected to a host computer. Figure
6.1 compares the acceleration time history measured by two sensing nodes and
reference sensor. The time history schemes of the wireless sensing nodes and
reference sensor have identical shapes and magnitudes. This confirms good-quality
data collection by the wireless sensing system. The same observation is made when
the signals are compared in terms of FFT (Fig. 6.2) and power spectral density (Fig.
6.3). Moreover, the peak of FFT indicates the wireless sensing node can identify the
shaking table is moving on 2 Hz Sin wave.
Figure 6.1 Acceleration record from wireless sensing nodes and reference sensor.
122
Figure 6.2 FFT of acceleration record from wireless sensing nodes and reference
sensor.
Figure 6.3 Power spectral density of acceleration record from wireless sensing
nodes and reference sensor.
Frequency(Hz)
0 1 2 3 4 5
Am
p.
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
Reference sensor
Wireless sensing node1
Wireless sensing node2
2D Graph 8
Frequency(Hz)
0 5 10 15 20 25 30
db
-140
-120
-100
-80
-60
-40
-20
0
Reference sensor
Wireless sensing node1
Wireless sensing node2
123
This experimental study then evaluated the feasibility and robustness of the
proposed system in a 1/8-scaled three-storey steel frame model (Fig.6.4) was
employed to verify the performance of the wireless sensing system placed on a
Quanser shaking table. Each floor weighed about 3.8 kg and each column had a
cross-sectional area of 80 mm2 and was 440 mm in height. Wireless sensing nodes
were deployed at the center of each floor. In this test, the model was excited by the
Quanser shaking table using earthquake time history excitation data. The
measurement period was 60 sec, during which data were oversampled. Oversampling,
typically used for anti-aliasing, was based on the Nyquist theorem. Measurement data
were sampled at 800Hz, and then down-sampled to a sampling rate of 200 Hz.
Measured structural dynamic responses were acquired by sensing nodes and sent to
the cluster head node connected to a host computer. Figure 6.5 shows the ground
acceleration time history measured by wireless sensing nodes. Following, acceleration
measurements from the wireless sensors identified natural frequencies by applying an
ANN-based system identification (ANNSI) model [46]. The corresponding first three
natural frequencies of the structure were 1.83, 5.42, and 7.96Hz. The modal
parameters can be considered reasonable by comparing to FEM model.
125
0 10 20 305 15 25
Time(sec)
-1
-0.6
-0.2
0.2
0.6
1
Acc
eler
atio
n(g)
input excitation
-1
-0.6
-0.2
0.2
0.6
1
Acc
eler
atio
n(g)
1st floor
-1
-0.6
-0.2
0.2
0.6
1
Acc
eler
atio
n(g)
2nd floor
-1
-0.6
-0.2
0.2
0.6
1
Acc
eler
atio
n(g)
3rd floor
Figure 6.5 Acceleration record from wireless sensing nodes
126
4 8 126 10 14
Time(s)
-1
-0.6
-0.2
0.2
0.6
1
no
malize
d r
esp
on
ses
1st floor
wireless
ANN
-1
-0.6
-0.2
0.2
0.6
1
no
malized
resp
on
ses
2nd floor
wireless
ANN
-1
-0.6
-0.2
0.2
0.6
1
no
malized
resp
on
ses
3rd floor
wireless
ANN
Figure 6.6. Response for ANN and wireless sensor node
127
6.2.2 Reliability of RF Communication Test
In a WSN, some packets may be lost during the RF communication. For instance,
packets may collide with each other if multiple nodes attempt to send packets to the
base station simultaneously. Packet loss may thus cause incomplete measurement
signals. Herein, the successful packet received rate (SPRR) was utilized to evaluate
the reliability of data transmission among nodes. The SPRR is defined as the number
of successfully received packets divided by the number of sent packets. In practice,
sensing nodes send a fixed number of packets to a base station. The base station can
thus calculate the SPRR by using the number of sent and received packets.
The SPRR was used to evaluate the packet sent interval. In this study, the packet
sent interval was 20 ms when the value of SPRR approached 1.0 whereas the SPRR
declined as the packet sent interval decreased during the wireless communication test.
In SHM, extended period measurement is necessary. In contrast with the
conventionally adopted wired sensor system, the wireless sensor node runs an
individual procedure on its CPU, with any exception event possibly disrupting node
operations. Therefore, the stability and robustness for extended period measurement
should be assessed. The measured period was set as 10, 30, 60 and 120 minutes,
respectively. The sampling rate was 200Hz. The sensor nodes sample and transmit
data to the base station via the proposed reliable data-sensing and transmission service.
The base station then calculates the SPRR for data loss evaluation. Figure 6.7
indicates that the average SPRR is almost 99% for 10, 30, 60 and 120 minutes. This
indicates that data loss was lower than 1% even in extended measuring period. It
means that the proposed system is robust and stable for long term monitoring.
128
Figure 6.7 SPRR for various period measurement
As mentioned earlier, packets may collide with each other if multiple nodes
attempt to send packets simultaneously. In this test, nodes with or without a
time-scheduling data transmission procedure are designated to send 100 packets to the
base station. The base station then calculates the SPRR to evaluate the data loss. With
the time-scheduling procedure, the SPRR is close to 100% (Fig. 6.8). Without
time-scheduling mechanism, if more than three nodes simultaneously send data to the
base station, then the SPRR dramatically decreases to 50% (Fig. 6.8). These analytical
results reveal that an effective time-scheduling procedure can increase the data
transmission integrity.
129
Figure 6.8 Comparison of success packets received rate(SPRR) for well Time
Schedule and Non Time Schedule.
Figure 6.9 presents the relation between RF power and distance and the SPRR.
Clearly, transmission range depends on RF power. A higher power output increases
range but also increases power consumption. Therefore, Figure 6.9 is a useful
reference when deploying the sensor nodes in the structure. Each node can be
allocated a different transmission power to optimize power consumption for varying
distance. In the experimental study, the farthest distance was 20m. Hence, power was
set to -15dbm to reduce power consumption.
130
Figure 6.9. Successful packet received rate (SPRR) to RF Power and Distance
6.2.3 Field Test in Bridge Structure
In Taiwan, the typhoon cause great damage on bridge due to Bridge Scour.
Hence, the monitoring of Bridge Scour has received considerable interest in the last
decade. Traditional site investigation methods are expensive and time-consuming and
may not always give a complete assessment. Geophysical methods can be used to
determine the riverbed profile below the water in a river. A trial of ground penetrating
radar (GPR) is a geophysical method that is particularly efficient in determining the
sub-bottom geological structure in a shallow freshwater environment. Forde et al.
[135] discussed a number of scour surveys using GPR. They concluded that GPR
131
surveys can be effective in determining both the water depth and sub-bottom
geological structure near bridge piers and abutments. Yankielun et al. [136] conducted
a laboratory experiment using time-domain reflectometry (TDR) for monitoring scour.
Their system capable of continuous round-the-clock operation can be used to
constantly monitor the extent of scour and can indicate changes in sediment depth of
less than 5 cm. The fiber Bragg grating (FBG) was also an alternative technique for
scour monitoring. Lin et al. [137] developed a FBG based monitoring system and
tested in the laboratory. Their FBG scour-monitoring systems can be used to measure
both the processes of scouring/deposition and the variations of water level. Several
experimental runs have been conducted in the scaled flume to demonstrate the
applicability of the system. Their experimental results indicate that the real-time FBG
monitoring system has the potential monitor the scour. However, further applications
in the field should be conducted. The robotic system was also presented for
underwater inspection of bridge substructures. DeVault [138] developed an automated
robotic system for underwater inspection of bridge substructures. This robotic system
can detect scour, deterioration, or damage to support columns by two identical mobile
robots designed to travel along opposite surfaces of the pier while connected to each
another by a cable and winch system. It provides positional data and sensor
information (video images) to the system operator for condition evaluation.
The conventional wired sensing system is expensive and infeasible.
Characterized by its low manufacturing costs, low power requirements, miniaturized
size, and no need for cabling, the wireless sensor networks (WSN) is an attractive
sensing technology for deploying sensors for bridge scour monitoring. Hence, the
developed integrated WSN-based SHM system was deployed in a bridge to test the
feasibility and practicality in real civil structure.
132
The proposed system was deploying in Nioudou Bridge. A field free vibration
test was conducted for obtaining dynamic response. Two piers were connected by a
steel bar for free vibration test. The soil ground was excavated to 4m depth to
simulate scouring. The free vibration test was performed twice for comparing intact
and scouring case. Dynamic responses were record and send to base station via
data-sensing and transmission service from wireless sensing nodes.
6.2.3.1 Experimental Setup
Figure 6.10 presents the experiment setup and sensor locations. The field
experiment was conducted in Nioudou Bridge. Two piers were connected by a steel
bar for a free vibration test. Hydraulic machinery was used to provide a tensile force
on the steel bar. The tensile force was increasing till the weak part of the steel bar was
broken. This broken energy forces the pier for free vibration. The maximum tensile
force was previously examined by a Tensile Test in the laboratory. Figure 6.11 shows
the maximum tensile force is 250kN. Three wireless sensing nodes were deployed on
the top, center and bottom of the pier 1(P1), respectively. Dynamic responses were
record and send to base station via reliable data-sensing and transmission service. The
soil ground was excavated to 4m depth to simulate scouring. The free vibration test
was performed twice for comparing intact and scouring case.
134
6.2.3.2 Experimental Results
Figure 6.12-6.15 shows the photograph of field experiment. Figure 6.12 shows
the Nioudou Bridge. The expansion joint was previously removed to increase the
displacement of pier when perform free vibration test. Figure 6.13 shows the weak
part of the steel bar. Two weak steel bars were used in intact and scouring case. The
steel bars were broken when the tensile force reach 180kN and 220kN respectively in
intact and scouring case. Figure 6.14 demonstrates the excavation condition to
simulate scouring. Figure 6.15 presents the location of wireless sensing node which is
fixed in the pier.
Figure 6.12 Photo of Nioudou Bridge
136
Figure 6.15 Wireless sensing node position.
In this experiment, the reliable data-sensing and transmission service which is
proposed in chapter 4 were simply modified to implement in this test. Figure 6.16
shows the data transportation and sensing process. First, the sensing modes were
deployed on the top, center and bottom of the pier. Next, after the sensing nodes and
cluster head node are initialized, the cluster head node sends an inquiring packet to
confirm whether the sensing nodes are ready. The cluster head node then broadcasted
packets to synchronize all sensor nodes and to remote sensing nodes for sensing;
sensors continued measuring and storing data in SDAM. The measurement period was
20 minute, during which data were sampled at 250Hz. Following the timeout, the
Imote2 nodes sent data to the base station under a time-scheduling mechanism to
avoid data collisions. Initially, the sensing nodes wait until a sending-delay time is
equivalent to (total sampling time + total packet sending time)* (node ID). For
instance, if the node ID is 0, the sensing node 0 transmits packet immediately since
137
the sending-delay equals zero. The data in each sensing node are then taken from the
array, filled in a packet, and sent to the base station. Based on this procedure, each
node sequentially sends the data to the base station with a respective sending-delay,
which can avoid the packet collision.
Figure 6.16 Modified reliable data-sensing and transmission service
138
Figure 6.17 shows the acceleration response measured by the sensing nodes. The
good-quality data collection by the wireless sensing system successfully records a
typical free vibration response of the pier. Figure 6.18 shows the responses on the top
sensing node for comparing the intact and scouring case. Observing to the response of
the free vibration, it indicates that the free vibration response for intact tend to stable
faster than the scouring case. Figure 6.19 compares two FFTs for intact and scouring
case. In frequency domain, the difference between intact and scouring case can be
observed. The work indicates that the proposed wireless sensing system was a reliable
technology for monitoring bridge structure.
Figure 6.17 Responses of sensor nodes for intact case.
139
Figure 6.18 Comparison of scouring and intact bridge pier with free vibration
response on the top node.
Figure 6.19 Comparison of FFT for scouring and intact bridge pie.
140
6.3 Experimental Study in Building Structure
This experimental study first exam the proposed damage detection approach in a
four-storey steel frame model (Fig. 6.20) placed on a Quanser shaking table. Each
floor weighed about 4 kg and each column had a cross-sectional area of 40 mm2 and
was 210 mm in height. Wireless sensing nodes were deployed at the center of each
floor. Four PZT patches are respectively bonded to the joint of the columns in 1st floor.
After measuring the several baseline impedance signatures and acceleration, damage
was introduced by completely loosening two bolts at joint 4 on first floor. In first
stage, the proposed SubFRFDI were used to detect the damaged floor in global state.
The identified result indicates that the proposed methods can detect the location of
damage on first floor, as shown in Fig. 6.21.
Figure 6.20 Experimental setup for wireless SHM system and impedance
measurement device
141
Figure 6.21 SubFRFDI value: damage on first floor.
After the global SHM approach, SubFRFDI, roughly indicate the location of
damaged floor. The EMI based damage detection approach then check the component
of building structure locally. The real part of impedance measurements of the PZT at
joint 1 to joint 4 are shown in Fig. 6.22. Figure 6.22 shows a baseline of undamaged
state and four damaged impedance curves. A damage metric chart is illustrated in Fig.
6.23. The damage metric chart based on RMSD is constructed to identify the local
damage position of structure. As shown in Fig. 6.23, the highest value identifies the
damage of joint 4 significantly.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
1 2 3 4
Su
bF
RF
DI
Number of Substructure
Damage on 1F
143
Figure 6.23 A damage metric chart of RMSD for damage on joint 4.
Subsequently, the proposed damage detection approach then test in a ¼ -scale
six-storey steel structure that was designed by the National Center for Research on
Earthquake Engineering (NCREE), Taiwan. Figure 6.24 shows an experimental setup
of the test structure. The floors, beams, and columns were connected using bolts.
Various measurement sensors, including 3-axes accelerometers, an LVDT, and
velocity sensors, were deployed. Simultaneously, the wireless SHM system was
deployed for sensing, logging, storing, processing, and analyzing thus obtained (Fig.
6.24). Four damage scenarios were considered to simulate states of damage to the
steel frame model. Table 6.1 and Fig. 6.25 show the four damage scenarios in which
the width and thickness of a column was reduced separately or simultaneously. All
experiments involved excitation using a 100 gal El Centro earthquake input on a
shaking table at NCREE.
144
Figure 6.24 A photograph of the test structure. (b) The Imote2.NET-based
sensing nodes fixed on the floor.
Table 6.1 Description damage scenarios for the steel-frame model.
Damage scenario
case
Description
Damage scenario 1 Reduced 3.75 cm width in the medium height of
each column at 1st floor.
Damage scenario 2 Reduced 7.5 cm width in the medium height of
each column at 1st floor.
Damage scenario 3 Reduced 7.5 cm width and 6mm thickness in the
medium height of each column at 1st floor.
Damage scenario 4 Reduced 12 cm width in the medium height of
each column at 3rd floor.
Sensing node
Ttest
structure
Test structure
145
Figure 6.25 Photographer of damage scenario: (a) damage scenario 1. (b)
damage scenario 2. (c) damage scenario 3. (d) damage scenario 4.
The proposed WSN-based SHM system was deployed on a steel frame model.
The measurement period was 60s, during which data were sampled at 200 Hz.
Besides using the proposed reliable data-sensing and transmission service, a
time-scheduling data transmission procedure was used to transfer raw data to the base
station. The reliable data-sensing and transmission service provides excellent data
sensing and transmission quality for determining the structural dynamic properties.
Measured dynamic responses were acquired by wireless sensing nodes and reference
sensors. Figure 6.26 compares the acceleration time history measured by sensing
nodes and reference sensor. Using the steel frame responses, modal properties were
146
identified by the system identification procedure. In Table 6.2, the modal parameters
identified using the ANN-based system identification approach [139] were compared
with the measured data obtained by wireless sensing nodes and conventional
reference sensors. It is confirmed that the data obtained using the wireless sensing
nodes possess good quality and can be utilized reasonably to calculate the modal
parameters. Figure 6.27 (a)-(f) show the FRF of structure for 1st floor to 6
th floor.
151
Table 6.2 Comparisons of modal parameters identified by wireless sensing nodes
and conventional reference sensors.
Frequency (Hz) Damping (%)
Mode Imote2 Conventional sensor RMSE Imote2 Conventional sensor RMSE
1 1.110 1.105 0.003 1.150 1.145 0.003
2 3.741 3.727 0.009 1.210 1.197 0.009
3 6.471 6.456 0.009 0.980 0.970 0.007
4 9.410 9.399 0.007 0.831 0.827 0.002
5 12.220 12.214 0.004 0.811 0.798 0.008
6 14.212 14.208 0.001 0.370 0.362 0.005
Figure 6.28 compares modal parameters identified using the Embedded FRF and
ANN-based system identification (ANNSI) [46] approach with the measured raw data
obtained by wireless sensing nodes. Comparisons demonstrate that the modal
parameters identified by Embedded FRF approximate those identified by ANNSI.
This comparison confirms that the proposed protocol is effective. Although the
proposed protocol is reasonable for wireless-based SHM, a performance evaluation of
embedded processing is essential. Table 6.3 shows the processing time of embedded
FFT and FRF for different data lengths. Practically, the numbers of FFT data points
was usually 1,024 or 2,048 points. The processing time for 1,024 points FFT and FRF
are 0.193s and 0.064s respectively. The same code was also implemented in 200Mhz
ARM processor-based .NET Micro Framework device. In this case, the processing
time for 1,024 points FFT and FRF necessitate cost 0.8s and 0.25s. Obviously, the
Imote2 with XScale processor and enhanced DSP core is sufficiently powerful for
processing data in SHM.
152
Figure 6.28 Comparisons of modal parameters identified by Embedded FRF and
ANNSI.
Table 6.3 The processing time of embedded FFT and FRF for different data
length.
Data length
512 1024 2048 4096
FFT 0.089s 0. 193s 0. 442s 0.970s
FRF 0.032s 0.064s 0.130s 0.259s
Total 0.121s 0.256s 0.572s 1.229s
Numerically simulated responses demonstrate that the proposed SubFRFDI can
be successfully used to identify the damage locations. The proposed SubFRFDI was
further applied to identifying damage locations in the experimental frame. Figure 6.29
shows the SubFRFDI for damage scenarios 1–4. The highest SubFRFDI value
indicates that the severest damage was located on the first floor in damage scenarios
1–3 (Figs. 6.29(a)– (c)). The SubFRFDI also indicates the damage location on the
third floor in damage scenario 4 (Fig. 6.29(d)).
154
Figure 6.29 The SubFRFDI for damage scenarios 1–4.
Subsequently, the EMI based damage detection approach then check the
component of building structure locally. Five PZT patches are respectively bonded to
the the columns in 1st floor as shown in Figure 6.30. The damage metric chart based
on RMSD is constructed to identify the local damage. As shown in Fig. 6.31-33, the
highest value locates the damage of location 5 significantly. As can be seen in the Fig
155
6.34, with an increase in extent of damage, there is a corresponding increase in the
damage metric values. Although the impedance method cannot exactly determine the
exact extent of the damage, an increasing damage metric with increased severity of
damage provides slightly quantitative information on the conditions of a structure.
Figure 6.30 PZT setup on the test structure.
1700 mm 1700 mm
1000 mm 1500 mm
6@
10
00
mm
X dir (North). Y dir (West).
Accelerometer
Displacement
Velocity
Strain Gage
S7 S6 S6 S9
Bolt1
Bolt2
Bolt3
Bolt4
PZT1 PZT2
PZT3 PZT4
PZT5
Crack
158
CHAPHER 7 CONCLUSIONS REMARKS
7.1 Conclusion
A novel framework of an integrated wireless sensor network-based structural
health monitoring system in buildings and civil infrastructures was developed in this
study. The architecture of this SHM system was presented. A global-local-integrated
damage detection approach was also developed for damage detection. Numerical and
experimental studies were conducted to verify the applicability of the proposed
approach. The following conclusions are drawn based on the integrated WSN-based
SHM system, global-local-integrated damage detection approach and experimental
results.
Integrated WSN-based SHM system
The Integrated WSN-based SHM system was developed in this study. This
system architecture consists of sensing nodes, cluster head nodes, transfer node, and
base station. Each sensing node is based on Imote2 platform which is comprised of a
microprocessor, sensor module, and communication device. The cluster head is a dule
core design which combine Imote2 platform and second embedded device. The
cluster head has extra wireless module and GPS. The base station is the highest level
end device that has largest memory, most powerful processor and highest
communication capability.
A new sensor board was developed. This sensor board integrates a six degrees of
freedom inertial measurement unit (6 DoF IMU) SD746 sensor chip and a GPS
module. In this system, a novel windmill-magnet integrated piezoelectric (WMIP)
159
energy harvesting system was also developed.
The proposed software system of integrated WSN-based SHM system is a
three-tier-framework, i.e., node, logging and processing tiers. The application
functions are developed in a Microsoft NETMF environment. The NETMF and
Crossbow’s API, i.e., application interface provided by manufacturing company of
Imote2, assist most libraries in dealing with all peripherals and drivers, allowing us to
develop individual applications for the SHM efficiently. The proposed user interface
shows how the sensor data can be represented in a Fig. chart and nodes distribution
can also be conveniently observed.
WSN-based middleware services for SHM application are developed. The
middleware services include reliable data-sensing and transmission, which are data
aggregation, reliable communication, and synchronized sensing. These middleware
services are implemented on the nodes running .NETMF and are generally applicable
to SHM application.
Global-local-integrated damage detection approach
This study developed a global-local-integrated damage detection approach. In the
first stage, the global approach detected changes in characteristics of the overall. After
identifying the damage and its approximate location, the second stage used local
detection method to identify the detailed conditions of the damage.
Substructure-based frequency response function and wavelet energy distribution
approaches are proposed for use in global damage detection. Local structural damage
is detected using an EMI-based method. Numerical results confirm the effectiveness
of the global-local-integrated damage detection approach for locating damage.
160
Experimental results
In experimental study, First off all, the integrated WSN-based SHM system is
verified involving sensor calibration, wireless communication quality, data losing,
interference in material, power consumption. The verification of sensing node
indicates the good-quality data collection by the wireless sensing system and the
modal parameters of structure can be correctly calculated by this system. In reliability
of RF communication test indicates that the proposed system is robust and stable for
long term monitoring via the proposed reliable data-sensing and transmission service.
This system was deployed in Nioudou Bridge. A field free vibration test was
conducted for obtaining dynamic response. The good-quality data collection by the
wireless sensing system successfully records a typical free vibration response of the
pier. By comparing two FFTs for intact and scouring case, in frequency domain, the
difference between intact and scouring case can be observed. The work indicates that
the proposed wireless sensing system was a reliable technology for monitoring Bridge
structure.
The proposed damage detection approach then was tested in a ¼ -scale six-storey
steel structure. Experimental analysis using global SHM approach shows that, both
SubFRFDI can indicates the approximate location of the damaged area, then the
EMI-based damage detection approach can check the components of the building
structure locally. The damage metric chart based on RMSD is constructed to identify
the position of local damage to the structure.
161
7.2 Future Study
Following the unfinished works and the drawbacks of the proposed approaches,
some potential directions on the future research are brought out.
In the proposed system, the main sensor on the sensing node is an accelerometer.
However, the resolution of this accelerometer can works well on detection large
vibration such as structural response excited by earthquake. For ambient vibration,
new sensor board should be developed in the future work.
This study proposed a novel windmill-magnet integrated piezoelectric (WMIP)
energy harvesting system. In the future work, the multiple energy harvesting methods
such as solar, piezoelectric and wind should be integrated. The energy harvesting
circuit needs to be included in the mote. Moreover, wireless power technology could
be considered a feasible plan for WSN in SHM.
This works developed a structure of class libraries which is branch out into three
basic class of NETMF, basic class library of Imote2, and SHM class. The SHM class,
a specialized class, is developed by this study for SHM application. However, in this
SHM class only FFT and FRF were included. In the future work, more embedded
signal processing class like wavelet should be developed.
Substructure-based frequency response function approaches are proposed for use
in global damage detection. However, these methods only focus on detecting and
locating damage. Further study should be conducted in deciding the extent of damage
in structure. EMI-based method was used to detect the local damage herein. In the
future work, the EMI-based sensor node should be developed and the local damage
index should be embedded in the node.
162
In experimental study, the developed integrated WSN-based SHM system was
deployed in a steel structure and in a bridge. The experimental result indicates this
system performs well either in Lab or in field test. In the future work, this system will
be considered deploying in different civil structure. When deploying in a large scaled
structure, the software and hardware of the system should be improved and modified.
An optimization approach can also be applied for deploying great amount sensor
nodes in the future works.
163
REFERENCE
1. Adams, D.E., Health monitoring of structural materials and components.
Methods with Applications. Willey, New York, 2007.
2. Worden, K., Farrar, C.R., Haywood, J., and Todd, M., A review of nonlinear
dynamics applications to structural health monitoring. Structural Controal and
Mealth Monitoring, 2008. 15(4): p. 540-567.
3. Sohn, H., Farrar, C.R., Hemez, F.M., Shunk, D.D., Stinemates, D.W., and
Nadler, B.R., A review of structural health monitoring literature: 1996-2001.
2004, Los Alamos National Laboratory.
4. Doebling, S.W., Farrar, C.R., and Prime, M.B., A summary review of
vibration-based damage identification methods. Shock and Vibration Digest,
1998. 30(2): p. 91-105.
5. Sohn, H. and Laboratory, L.A.N., A review of structural health monitoring
literature: 1996-2001. 2004: Los Alamos National Laboratory.
6. Inman, D.J., Damage prognosis for aerospace, civil and mechanical systems.
2005: John Wiley & Sons Inc.
7. Ansari, F., Sensing issues in civil structural health monitoring. 2005: Kluwer
Academic Pub.
8. Chang, P., Flatau, A., and Liu, S., Review paper: health monitoring of civil
infrastructure. Structural Health Monitoring, 2003. 2(3): p. 257.
9. Farrar, C., Historical overview of structural health monitoring, in Lecture notes
on structural health monitoring using statistical pattern recognition. 2001, Los
Alamos Dynamics, Los Alamos, NM.
10. Celebi, M., Seismic instrumentation of buildings (with emphasis on federal
buildings), in Special GSA/USGS Project, an administrative report. 2002,
United States Geological Survey: Menlo Park, CA.
11. Nagayama, T., Sim, S., Miyamori, Y., and Spencer Jr, B., Issues in structural
health monitoring employing smart sensors. Smart Structures and Systems,
2007. 3(3): p. 299-320.
12. Lynch, J.P. and Loh, K.J., A summary review of wireless sensors and sensor
networks for structural health monitoring. Shock and Vibration Digest, 2006.
164
38(2): p. 91-130.
13. Spencer Jr, B.F., Ruiz-Sandoval, M.E., and Kurata, N., Smart sensing
technology: opportunities and challenges. Structural Control & Health
Monitoring, 2004. 11(4): p. 349-368.
14. Chavali, M., Lin, T.H., Wu, R.J., Luk, H.N., and Hung, S.L., Active 433
MHz-W UHF RF-powered chip integrated with a nanocomposite
m-MWCNT/polypyrrole sensor for wireless monitoring of volatile anesthetic
agent sevoflurane. Sensors and Actuators a-Physical, 2008. 141(1): p. 109-119.
15. Xu, N., Rangwala, S., Chintalapudi, K., Ganesan, D., Broad, A., Govindan, R.,
and Estrin, D. A wireless sensor network for structural monitoring. Proceeding
of 2nd international conference on Embedded networked sensor systems.
2004.
16. Yuan, S.F., Lai, X.S., Zhao, X., Xu, X., and Zhang, L., Distributed structural
health monitoring system based on smart wireless sensor and multi-agent
technology. Smart Materials & Structures, 2006. 15(1): p. 1-8.
17. Kurata, N., Spencer Jr, B., Ruiz-Sandoval, M., Miyamoto, Y., and Sako, Y., A
study on building risk monitoring using wireless sensor network MICA mote.
Strain, 2003. 1: p. 35.
18. Kim, S., Pakzad, S., Culler, D., Demmel, J., Fenves, G., Glaser, S., and Turon,
M. Health monitoring of civil infrastructures using wireless sensor networks.
Proceeding of 6th international conference on Information processing in
sensor networks. 2007.
19. Doebling, S.W., Farrar, C.R., and Goodman, R.S. Effects of measurement
statistics on the detection of damage in the Alamosa Canyon Bridge. Proceeding
of Proceedings 15th International Modal Analysis Conference. 1997.
20. Farrar, C.R., Doebling, S.W., Cornwell, P.J., and Straser, E.G. Variability of
modal parameters measured on the Alamosa Canyon Bridge. Proceeding of
Proceedings 15th International Modal Analysis Conference. 1997.
21. Doebling, S.W., Farrar, C.R., Prime, M.B., and Shevitz, D.W., Damage
identification and health monitoring of structural and mechanical systems from
changes in their vibration characteristics: a literature review. 1996, Los Alamos
National Lab., NM (United States).
22. Salawu, O., Detection of structural damage through changes in frequency: a
165
review. Engineering Structures, 1997. 19(9): p. 718-723.
23. West, W.M. Illustration of the use of modal assurance criterion to detect
structural changes in an orbiter test specimen. Proceeding of Air Force
Conference on Aircraft Structural Integrity. 1986.
24. Mayes, R., Error localization using mode shapes: An application to a two link
robot arm, in 10th International Modal Analysis Conference. 1991, Sandia
National Labs., Albuquerque, NM (United States): Albuquerque, NM.
25. Kam, T. and Lee, T., Detection of cracks in structures using modal test data.
Engineering fracture mechanics, 1992. 42(2): p. 381-387.
26. Ko, J., Wong, C., and Lam, H. Damage detection in steel framed structures by
vibration measurement approach. Proceeding of 12th International Modal
Analysis Conference. 1994.
27. Farrar, C.R. and Doebling, S.W. An overview of modal-based damage
identification methods. Proceeding of DAMAS Conference. 1997.
28. Fox, C. The location of defects in structures-A comparison of the use of natural
frequency and mode shape data. Proceeding of the 10th International Modal
Analysis Conference. 1992.
29. Pandey, A., Biswas, M., and Samman, M., Damage detection from changes in
curvature mode shapes. Journal of Sound and Vibration, 1991. 145(2): p.
321-332.
30. Topole, K.G. and Stubbs, N., Non destructive damage evaluation of a structure
from limited modal parameters. Earthquake Engineering & Structural
Dynamics, 1995. 24(11): p. 1427-1436.
31. Chance, J., Tomlinson, G., and Worden, K. A simplified approach to the
numerical and experimental modelling of the dynamics of a cracked beam.
Proceedings of the 12th International Modal Analysis Conference. 1994.
32. Chen, J.C. and Garbat, J.A., Analytical model improvement using modal test
results. Aiaa Journal, 1980. 18: p. 684-690.
33. SMITH, S. Iterative use of direct matrix updates- Connectivity and convergence.
Proceeding of 33rd AIAA Structures, Structural Dynamics and Materials
Conference. 1992.
34. Liu, P.L., Identification and damage detection of trusses using modal data.
166
Journal of Structural Engineering, 1995. 121(4): p. 599-608.
35. KAOUK, M. and Zimmerman, D.C., Structural damage assessment using a
generalized minimum rank perturbation theory. Aiaa Journal, 1994. 32(4): p.
836-842.
36. Hemez, F. and Farhat, C., Structural damage detection via a finite element
model updating methodology. MODAL ANALYSIS-The International Journal
of Analytical and Experimental Modal Analysis, 1995. 10(3): p. 152-166.
37. Norris, M.A. and Meirovitch, L., On the problem of modelling for parameter
identification in distributed structures. International Journal for Numerical
Methods in Engineering, 1989. 28(10): p. 2451-2463.
38. Haug, E. and Choi, K., Structural design sensitivity analysis with generalized
global stiffness and mass matrices. Aiaa Journal, 1984. 22(9): p. 1299-1303.
39. Ricles, J. and Kosmatka, J., Damage detection in elastic structures using
vibratory residual forces and weighted sensitivity. Aiaa Journal, 1992. 30(9): p.
2310-2316.
40. Lim, T.W. and Kashangaki, T.A.L., Structural damage detection of space truss
structures using best achievable eigenvectors. Aiaa Journal, 1994. 32: p.
1049-1057.
41. Zimmerman, D.C. and MOHAMED, K., Eigenstructure assignment approach
for structural damage detection. Aiaa Journal, 1992. 30(7): p. 1848-1855.
42. Sohn, H., Farrar, C.R., Hunter, N.F., and Worden, K., Structural Health
Monitoring Using Statistical Pattern Recognition Techniques. Journal of
Dynamic Systems, Measurement, and Control, 2001. 123(4): p. 706-711.
43. Lei, Y., Kiremidjian, A., Nair, K., Lynch, J., Law, K., Kenny, T., Carryer, E.,
and Kottapalli, A. Statistical damage detection using time series analysis on a
structural health monitoring benchmark problem. Proceeding of 9th
International Conference on Applications of Statistics and Probability in Civil
Engineering. 2003.
44. Chen, H., Qi, G., Yang, J., and Amini, F., Neural network for structural
dynamic model identification. Journal of engineering mechanics, 1995. 121: p.
1377.
45. Wu, X., Ghaboussi, J., and Garrett, J., Use of neural networks in detection of
structural damage. Computers & Structures, 1992. 42(4): p. 649-659.
167
46. Wen, C.M., Hung, S.L., Huang, C.S., and Jan, J.C., Unsupervised fuzzy neural
networks for damage detection of structures. Structural Controal and Mealth
Monitoring, 2007. 14(1): p. 144-161.
47. Kudva, J., Munir, N., and Tan, P., Damage detection in smart structures using
neural networks and finite-element analyses. Smart Materials and Structures,
1992. 1: p. 108.
48. Poudel, U., Fu, G., and Ye, J., Structural damage detection using digital video
imaging technique and wavelet transformation. Journal of Sound and
Vibration, 2005. 286(4-5): p. 869-895.
49. Kim, H. and Melhem, H., Damage detection of structures by wavelet analysis.
Engineering Structures, 2004. 26(3): p. 347-362.
50. Ovanesova, A.V. and Suárez, L.E., Applications of wavelet transforms to
damage detection in frame structures. Engineering Structures, 2004. 26(1): p.
39-49.
51. Huang, C.S. and Su, W.C., Identification of modal parameters of a time
invariant linear system by continuous wavelet transformation. Mechanical
Systems and Signal Processing, 2007. 21(4): p. 1642-1664.
52. Huang, C.S., Hung, S.L., Lin, C.I., and Su, W.C., A wavelet-based approach to
identifying structural modal parameters from seismic response and free
vibration data. Computer-Aided Civil and Infrastructure Engineering, 2005.
20(6): p. 408-423.
53. Wu, N. and Wang, Q., Experimental studies on damage detection of beam
structures with wavelet transform. International Journal of Engineering
Science, 2011. 49(3): p. 253-261.
54. Han, J.-G., Ren, W.-X., and Sun, Z.-S., Wavelet packet based damage
identification of beam structures. International Journal of Solids and
Structures, 2005. 42(26): p. 6610-6627.
55. Poornachandra, S., Wavelet-based denoising using subband dependent
threshold for ECG signals. Digital Signal Processing, 2008. 18(1): p. 49-55.
56. Pasti, L., Walczak, B., Massart, D.L., and Reschiglian, P., Optimization of
signal denoising in discrete wavelet transform. Chemometrics and Intelligent
Laboratory Systems, 1999. 48(1): p. 21-34.
57. Lee, B. and Tarng, Y., Drill fracture detection by the discrete wavelet transform.
168
Journal of Materials Processing Technology, 2000. 99(1-3): p. 250-254.
58. Prabhakar, S., Mohanty, A., and Sekhar, A., Application of discrete wavelet
transform for detection of ball bearing race faults. Tribology International,
2002. 35(12): p. 793-800.
59. Sansalone, M., Carino, N.J., and Division, C.f.B.T.S., Impact-echo: a method
for flaw detection in concrete using transient stress waves. 1986: US Dept. of
Commerce, National Bureau of Standards, Center for Building Technology,
Structures Division.
60. Davis, A.G., Evans, J.G., and Hertlein, B.H., Nondestructive evaluation of
concrete radioactive waste tanks. Journal of performance of constructed
facilities, 1997. 11: p. 161.
61. Chady, T., Enokizono, M., and Sikora, R., Crack detection and recognition
using an eddy current differential probe. Magnetics, IEEE Transactions on,
1999. 35(3): p. 1849-1852.
62. Tian, G.Y., Sophian, A., Taylor, D., and Rudlin, J., Multiple sensors on pulsed
eddy-current detection for 3-D subsurface crack assessment. Sensors Journal,
IEEE, 2005. 5(1): p. 90-96.
63. Washer, G.A. Detection of fatigue cracks in eyebars using time of flight
diffraction. Proceeding of Conf. on StructuralMaterials Technology and
Nondestructive Testing. 1996.
64. Liang, C., Sun, F., and Rogers, C., Coupled Electro-Mechanical Analysis of
Adaptive Material Systems--Determination of the Actuator Power Consumption
and System Energy Transfer. Journal of Intelligent Material Systems and
Structures, 1994. 5(1): p. 12.
65. Sun, F., Rogers, C.A., and Liang, C. Structural frequency response function
acquisition via electric impedance measurement of surface-bonded piezoelectric
sensor/actuator. Proceeding of Structures, Structural Dynamics, and
Materials Conference. 1995.
66. Chaudhry, Z., Lalande, F., Ganino, A., and Rogers, C. Monitoring the integrity
of composite patch structural repair via piezoelectric actuators/sensors.
Proceeding of 36th AIAA/ASME/ASCE/AHS/ASC Structures, Structural
Dynamics, and Materials Conference. 1995.
67. Park, G., Kabeya, K., Cudney, H.H., and Inman, D.J., Impedance-based
169
structural health monitoring for temperature varying applications. JSME
international journal. Series A, Solid mechanics and material engineering,
1999. 42(2): p. 249-258.
68. Park, G., Cudney, H.H., and Inman, D.J., Impedance-based health monitoring
of civil structural components. Journal of Infrastructure Systems, 2000. 6(4): p.
153-160.
69. Yang, Y., Hu, Y., and Lu, Y., Sensitivity of PZT impedance sensors for damage
detection of concrete structures. Sensors, 2008. 8(1): p. 327-346.
70. Kim, J., Vibration and impedance monitoring for prestress-loss prediction in
PSC girder bridges. Smart Structures and Systems, 2009. 5(1): p. 81-94.
71. Park, G., Sohn, H., Farrar, C.R., and Inman, D.J., Overview of piezoelectric
impedance-based health monitoring and path forward. Shock and Vibration
Digest, 2003. 35(6): p. 451-464.
72. Mascarenas, D., Flynn, E., Todd, M., Park, G., and Farrar, C., Wireless Sensor
Technologies for Monitoring Civil Structures. Sound and Vibration, 2008. 42(4):
p. 16.
73. Kim, S., Pakzad, S., Culler, D., Demmel, J., Fenves, G., Glaser, S., and Turon,
M. Health monitoring of civil infrastructures using wireless sensor networks.
Proceeding of 6th international conference on Information processing in
sensor networks 2007.
74. Cho, S., Jang, S.A., Jo, H., Mechitov, K., Rice, J.A., Jung, H.J., Yun, C.B.,
Spencer Jr, B.F., Nagayama, T., and Seo, J. Structural health monitoring
system of a cable-stayed bridge using a dense array of scalable smart sensor
network. Proceeding of Sensors and Smart Structures Technologies for Civil,
Mechanical, and Aerospace Systems. 2010.
75. Hoult, N.A., Fidler, P.R.A., Hill, P.G., and Middleton, C.R., Long-term wireless
structural health monitoring of the Ferriby Road Bridge. Journal of Bridge
Engineering, 2010. 15: p. 153.
76. Kim, J., Lynch, J.P., Lee, J.J., and Lee, C.G., Truck-based mobile wireless
sensor networks for the experimental observation of vehicle¡Vbridge interaction.
Smart Materials and Structures, 2011. 20: p. 065009.
77. Castillo-Effer, M., Quintela, D.H., Moreno, W., Jordan, R., and Westhoff, W.
Wireless sensor networks for flash-flood alerting. Proceeding of the Fifth IEEE
170
International Caracas Conference on Devices. 2004.
78. Simon, G., Maroti, M., Ledeczi, A., Balogh, G., Kusy, B., Nadas, A., Pap, G.,
Sallai, J., and Frampton, K. Sensor network-based countersniper system.
Proceeding of the 2nd international conference on Embedded networked
sensor systems 2004.
79. Werner-Allen, G., Lorincz, K., Welsh, M., Marcillo, O., Johnson, J., Ruiz, M.,
and Lees, J., Deploying a wireless sensor network on an active volcano. Ieee
Internet Computing, 2006: p. 18-25.
80. Akyildiz, I.F., Su, W., Sankarasubramaniam, Y., and Cayirci, E., Wireless
sensor networks: a survey. Computer Networks-the International Journal of
Computer and Telecommunications Networking, 2002. 38(4): p. 393-422.
81. Yick, J., Mukherjee, B., and Ghosal, D., Wireless sensor network survey.
Computer Networks, 2008. 52(12): p. 2292-2330.
82. Younis, M. and Akkaya, K., Strategies and techniques for node placement in
wireless sensor networks: A survey. Ad Hoc Networks, 2008. 6(4): p. 621-655.
83. Nagayama, T., Spencer Jr, B.F., Mechitov, K.A., and Agha, G.A., Middleware
services for structural health monitoring using smart sensors. Smart Structures
and Systems, 2009. 5(2): p. 119-137.
84. Lynch, J.P., An overview of wireless structural health monitoring for civil
structures. Philosophical Transactions of the Royal Society a-Mathematical
Physical and Engineering Sciences, 2007. 365(1851): p. 345-372.
85. Cho, S., Yun, C.B., Lynch, J.P., Zimmerman, A.T., Spencer, B.F., and
Nagayama, T., Smart Wireless Sensor Technology for Structural Health
Monitoring of Civil Structures. International Journal of Steel Structures, 2008.
8(4): p. 267-275.
86. Paek, J., Chintalapudi, K., Govindan, R., Caffrey, J., and Masri, S. A wireless
sensor network for structural health monitoring: Performance and experience.
Proceeding of The Second IEEE Workshop on Embedded Networked Sensors
(EmNetS-II). 2005.
87. Straser, E.G., Kiremidjian, A.S., and Center, J.A.B.E.E., A modular, wireless
damage monitoring system for structures. 1998: The John A. Blume
Earthquake Engineering Center.
88. Lynch, J.P., Law, K.H., Kiremidjian, A.S., Kenny, T.W., Carryer, E., and
171
Partridge, A. The design of a wireless sensing unit for structural health
monitoring. Proceeding of 3rd Int’l Ws. Structural Health Monitoring. 2001.
89. Wang, Y., Lynch, J.P., and Law, K.H., Validation of an integrated network
system for real-time wireless monitoring of civil structures. Structural health
monitoring, 2005: advancements and challenges for implementation, 2005.
1050(48109): p. 275.
90. Aoki, S., Fujino, Y., and Abe, M. Intelligent bridge maintenance system using
MEMS and network technology. Proceeding of SPIE Conf. 2003.
91. Chung, H.C., Enomoto, T., Shinozuka, M., Chou, P., Park, C., Yokoi, I., and
Morishita, S. Real-time visualization of structural response with wireless
MEMS sensors. Proceeding of 13th world conf. Earthquake Engineering.
2004.
92. Farrar, C.R., Allen, D.W., Park, G., Ball, S., and Masquelier, M.P., Coupling
sensing hardware with data interrogation software for structural health
monitoring. Shock and Vibration, 2006. 13(4): p. 519-530.
93. Chen, B. and Liu, W., Mobile agent computing paradigm for building a flexible
structural health monitoring sensor network. Computer Aided Civil and
Infrastructure Engineering, 2010. 25(7): p. 504-516.
94. Xu, N., Rangwala, S., Chintalapudi, K.K., Ganesan, D., Broad, A., Govindan,
R., and Estrin, D., A wireless sensor network For structural monitoring.
Proceedings of the 2nd international conference on Embedded networked
sensor systems, 2004: p. 13-24.
95. ShenfangYuan, X.L., Zhao, X., Xu, X., and Zhang, L., Distributed structural
health monitoring system based on smart wireless sensor and multi-agent
technology. Smart Mater. Struct, 2006. 15: p. 1-8.
96. Kurata, N., Spencer Jr, B.F., Ruiz-Sandoval, M., Miyamoto, Y., and Sako, Y.,
A study on building risk monitoring using wireless sensor network MICA mote.
Strain. 1: p. 35.
97. Hoblos, G., Staroswiecki, M., and Aitouche, A. Optimal design of fault tolerant
sensor networks. Proceeding of IEEE International Conference on Control
Applications. 2000.
98. Cho, S.H. and Chandrakasan, A.P. Energy efficient protocols for low duty cycle
wireless microsensor networks. Proceeding of 33rd Annual Hawaii
172
International Conference on System Sciences. 2001.
99. Rabaey, J.M., Ammer, M.J., da Silva Jr, J.L., Patel, D., and Roundy, S.,
PicoRadio supports ad hoc ultra-low power wireless networking. Computer,
2000. 33(7): p. 42-48.
100. Intanagonwiwat, C., Govindan, R., and Estrin, D. Directed diffusion: A
scalable and robust communication paradigm for sensor networks. Proceeding
of ACM Mobi-Com'00. 2000.
101. Kahn, J.M., Katz, R.H., and Pister, K.S.J. Next century challenges: mobile
networking for smart dust. Proceeding of ACM MobiCom’99. 1999.
102. Alliance, Z.B., ZigBee Specifications Version 1.0., San Ramon, CA, USA,
December. 2004. 2004.
103. Gutierrez, J.A., Naeve, M., Callaway, E., Bourgeois, M., Mitter, V., and Heile,
B., IEEE 802.15. 4: a developing standard for low-power low-cost wireless
personal area networks. Network, IEEE, 2001. 15(5): p. 12-19.
104. Woo, A., Madden, S., and Govindan, R., Networking support for query
processing in sensor networks. Communications of the Acm, 2004. 47(6): p.
47-52.
105. Kopetz, H., Damm, A., Koza, C., Mulazzani, M., Schwabl, W., Senft, C., and
Zainlinger, R., Distributed fault-tolerant real-time systems: The Mars approach.
Micro, IEEE, 1989. 9(1): p. 25-40.
106.Horauer, M., Schossmaier, K., Schmid, U., Holler, R., Kero, N., and ., V.U.O.T.
PSynUTC-Evaluation of a High-Precision Time Synchronization Prototype
System for Ethernet LANs. Proceeding of 34th Annual Precise Time and Time
Interval Meeting (PTTI). 2002.
107. Mills, D.L., Internet time synchronization: The network time protocol.
Communications, IEEE Transactions on, 1991. 39(10): p. 1482-1493.
108. Syed, A. and Heidemann, J. Time synchronization for high latency acoustic
networks. Proceeding of IEEE Infocom. 2006.
109. Nagayama, T., Spencer, B.F., and Rice, J.A., Autonomous decentralized
structural health monitoring using smart sensors. Structural Control & Health
Monitoring, 2009. 16(7-8): p. 842-859.
110. Lanzara, G., Feng, J., and Chang, F.K., Design of micro-scale highly
173
expandable networks of polymer-based substrates for macro-scale applications.
Smart Materials & Structures, 2010. 19(4): p. 045013.
111. Gratzel, M., Solar energy conversion by dye-sensitized photovoltaic cells. Inorg.
Chem, 2005. 44(20): p. 6841-6851.
112. Sodano, H., Simmers, G., Dereux, R., and Inman, D., Recharging batteries
using energy harvested from thermal gradients. Journal of Intelligent Material
Systems and Structures, 2007. 18(1): p. 3.
113. Beeby, S., Tudor, M., and White, N., Energy harvesting vibration sources for
microsystems applications. Measurement Science and Technology, 2006. 17: p.
R175.
114. Ottman, G., Hofmann, H., Bhatt, A., and Lesieutre, G., Adaptive piezoelectric
energy harvesting circuit for wireless remote power supply. IEEE Transactions
on Power Electronics, 2002. 17(5): p. 669-676.
115. Ganeriwal, S., Kumar, R., and Srivastava, M. Timing-sync protocol for sensor
networks. Proceeding of 1st international conference on Embedded networked
sensor systems. 2003.
116. Cauberghe, B., Guillaume, P., Verboven, P., Vanlanduit, S., and Parloo, E.,
Frequency response function-based parameter identification from short data
sequences. Mechanical Systems and Signal Processing, 2004. 18(5): p.
1097-1116.
117. Celic, D. and Boltezar, M., Identification of the dynamic properties of joints
using frequency-response functions. Journal of Sound and Vibration, 2008.
317(1-2): p. 158-174.
118. Wang, Z., Lin, R.M., and Lim, M.K., Structural damage detection using
measured FRF data. Computer Methods in Applied Mechanics and
Engineering, 1997. 147(1-2): p. 187-197.
119. Lee, U. and Shin, J., A frequency-domain method of structural damage
identification formulated from the dynamic stiffness equation of motion.
Journal of Sound and Vibration, 2002. 257(4): p. 615-634.
120. Thyagarajan, S.K., Schulz, M.J., Pai, P.F., and Chung, J., Detecting structural
damage using frequency response functions. Journal of Sound and Vibration,
1998. 210(1): p. 162-170.
121. Lee, U. and Shin, J., A frequency response function-based structural damage
174
identification method. Computers and Structures, 2002. 80(2): p. 117-132.
122. Sampaio, R.P.C., Maia, N.M.M., and Silva, J.M.M., Damage detection using
the frequency-response-function curvature method. Journal of Sound and
Vibration, 1999. 226(5): p. 1029-1042.
123. Maia, N.M.M., Silva, J.M.M., Almas, E.A.M., and Sampaio, R.P.C., Damage
detection in structures: from mode shape to frequency response function
methods. Mechanical Systems and Signal Processing, 2003. 17(3): p. 489-498.
124. Liu, X., Lieven, N.A.J., and Escamilla-Ambrosio, P.J., Frequency response
function shape-based methods for structural damage localisation. Mechanical
Systems and Signal Processing, 2009. 23(4): p. 1243-1259.
125. Ni, Y.Q., Zhou, X.T., and Ko, J.M., Experimental investigation of seismic
damage identification using PCA-compressed frequency response functions and
neural networks. Journal of Sound and Vibration, 2006. 290(1-2): p. 242-263.
126. Furukawa, A., Otsuka, H., and Kiyono, J., Structural damage detection
method using uncertain frequency response functions. Computer-Aided Civil
and Infrastructure Engineering, 2006. 21(4): p. 292-305.
127. Kanwar, V.S., Kwatra, N., Aggarwal, P., and Gambir, M.L., Health
monitoring of RCC building model experimentally and its analytical validation.
Engineering Computations, 2008. 25(7-8): p. 677-693.
128. Hsu, T. and Loh, C. Damage detection using frequency response functions
under ground excitation. Proceeding of Sensors and Smart Structures
Technologies for Civil, Mechanical, and Aerospace Systems. 2009.
129. Park, G. and Inman, D., Structural health monitoring using piezoelectric
impedance measurements. Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences, 2007. 365(1851): p. 373.
130. Liang, C., Sun, F., and Rogers, C., Coupled electro-mechanical analysis of
adaptive material systems¡Xdetermination of the actuator power consumption
and system energy transfer. Journal of Intelligent Material Systems and
Structures, 1994. 5(1): p. 12.
131. Sun, F., Chaudhry, Z., Liang, C., and Rogers, C., Truss structure integrity
identification using PZT sensor-actuator. Journal of Intelligent Material
Systems and Structures, 1995. 6(1): p. 134-139.
132. Peairs, D.M., Park, G., and Inman, D.J., Improving accessibility of the
175
impedance-based structural health monitoring method. Journal of Intelligent
Material Systems and Structures, 2004. 15(2): p. 129.
133. Nokes, J. and Cloud, G., The application of interferometric techniques to the
nondestructive inspection of fiber-reinforced materials. Experimental
Mechanics, 1993. 33(4): p. 314-319.
134. Esteban, J., Modeling of the sensing region of a piezoelectric actuator/sensor.
1996, Ph. D. Dissertation, Virginia Polytechnic Institute and State University,
Blacksburg, VA.
135. Forde, M., McCann, D., Clark, M., Broughton, K., Fenning, P., and Brown,
A., Radar measurement of bridge scour. Ndt & E International, 1999. 32(8): p.
481-492.
136. Yankielun, N. and Zabilansky, L., Laboratory investigation of time-domain
reflectometry system for monitoring bridge scour. Journal of Hydraulic
engineering, 1999. 125(12): p. 1279-1284.
137. Lin, Y., Chen, J., Chang, K., Chern, J., and Lai, J., Real-time monitoring of
local scour by using fiber Bragg grating sensors. Smart Materials and
Structures, 2005. 14: p. 664.
138. DeVault, J., Robotic system for underwater inspection of bridge piers.
Instrumentation & Measurement Magazine, IEEE, 2002. 3(3): p. 32-37.
139. Huang, C.S., Hung, S.L., Wen, C.M., and Tu, T.T., A neural network approach
for structural identification and diagnosis of a building from seismic response
data. Earthquake Engineering & Structural Dynamics, 2003. 32(2): p.
187-206.
176
Vita
林子軒 (Tzu-Hsuan Lin)
個人資料
生日:1978/06/05
性別:男性
住址:新竹市東大路二段172巷1弄7號1樓
電話:03-5712121#54965, 0963354403
Email: [email protected]
兵役:役畢
現職
國立交通大學土木工程系博士候選人(指導教授:洪士林教授)
國立交通大學電機與控制工程研究所碩士班(指導教授:廖德誠教授)
學歷
2009.7~2009.9 東京大學訪問學生(Host: Professor Yozo Fujino )
2005.9 ~Now 國立交通大學土木工程研究所博士班結構組
2007.6~ Now 雙主修國立交通大學電機與控制工程研究所碩士班
2001.9 ~ 2003.6 國立交通大學土木工程研究所碩士
2001.9 ~ 2003.6 國立交通大學資訊管理研究所輔所
1997.9 ~ 2001.6 國立台北科技大學土木系(第一名畢業)
榮譽
2009.7~9 日本交流協會暑期交換學生獎學金(兩個月共,72 萬日幣)赴
東京大學研究,並獲選為得獎代表,上台致謝。(全國 18 名,
唯一的土木領域)
2010 中華顧問工程司,工程科技獎學金,博士班獎學金 10 萬元。
2007~2009 中興工程研究發展基金會博士班獎助學金 (土木領域全國最
高獎學金,每年 30 萬元),連續三年獲獎
2006 中興工程顧問社優秀學生獎
177
研究成果目錄:論文及著述
A. 學位論文
林子軒,‖土木工程結構修復補強知識管理與支援決策輔助系統之研究” 碩士論
文,國立交通大學土木工程系,2003。
B. 期刊論文
[1] T. H. Lin, S. L. Hung, Y. C. Lu, T. Nagayama and Y. Fujino (2011),
Development of An Embedded Damage Detection Approach Using
Imote2.net-Based Wireless Structural Health Monitoring, Structural health
monitoring-an international journal. (Reviewing)
[2] T. H. Lin, S. L. Hung, C. S. Huang, and T. K. Lin (2012), Detection of damage
location using a novel substructure-based frequency response function approach
with a wireless sensing system, International Journal of Structural Stability and
Dynamics, has been accepted for publication in Vol. 12, No. 3 (June 2012)
IF:0.721
[3] T. H. Lin, S. L. Hung, M. Chavali, R. J. Wu, H. N. Luk and Q. Fei (2010),
Towards Development of Wireless Sensor System For Monitoring Anesthetic
Agents, Sensor Letters. 8, 767-776. IF:0.626
[4] R. J. Wu, Y. C. Huang, M. R. Yu, T. H. Lin, S. L. Hung (2008), Application of
m-CNTs/NaClO4/Ppy to A Fast Response, Room Working Temperature Ethanol
Sensor‖. Sensors and Actuators B: Chemical, 134(1), 213-218. IF:3.083
[5] Murthy Chavali , T. H. Lin, R.J. Wu, H.N. Luk and S.L. Hung (2008), ―Active
433 MHz-W UHF RF powered chip integrated with a nano composite
m-MWCNT/Polypyrrole sensor for wireless monitoring of anesthetic agent
Sevoflurane‖, Sensors & Actuators: A. Physical, 141(1), 109-119 IF:1.674
[6] R.J. Wu, Y.C. Huang, M. Chavali, T.H. Lin, S.L. Hung and H.N. Liud (2007),
―New sensing technology for detection of the common inhalational anesthetic
agent sevoflurane using conducting polypyrrole films‖, Sensors & Actuators: B.
Chemical 126 (2), 387-393. IF:3.083
[7] 吳仁彰, 余明儒, 林子軒, 洪士林 (2008/12/1)。CuO-ZnO 材料於半導
體型快速酒精感測之研究。 科儀新知,第 30 卷,第 3 期,第 167 輯,頁
48 -55 。
[8] 吳仁彰, 黃裕清, 林子軒, 龍彥先, 吳登峻, 梁元豪 (2008/5/1)。
CNT/NaClO4/Ppy 材料於快速酒精感測之研究。 量測資訊,第 121 期,頁
58-63。
[9] 林子軒、洪士林、吳仁彰、黃裕清、陸翔寧、柴梅熙,「無線感測網路應用
178
於醫院建築環境之監測」,科儀新知,第 28 卷,第 2 期,76–84 頁,民
國 95 年 10 月。
[10] 吳仁彰, 黃裕清, 陸翔寧, 林子軒, 洪士林, 柴梅熙,「麻醉氣體七氟烷的量
測技術研究」,科儀新知,第 27 卷,第 6 期,87–93 頁(2006)。
[11] 李有豐、刁健原、林子軒,「碳纖維強化高分子複合材料(CFRP)補強建築
物 RC 構件簡易查圖法」,結構工程,第十七卷,第一期,第 43-60 頁(2002)。
179
C.研討會論文
[1] T. H. Lin, Y. C. Lu and S. L. Hung, Field Study of Bridge Scour Using
Imote2.NET‐Based Wireless Monitoring System, 6th International Workshop
on Advanced Smart Materials and Smart Structures Technology, Dalian, China,
July 25‐26, 2011.(Accepted)
[2] C. S. Hsieh, D. C. Liaw and T. H. Lin, Optimal Filtering Method to Structural
Damage Estimation Under Ground Excitation (I): Kalman Filtering Approach,
The 8th Asian Control Conference (ASCC 2011), Kaohsiung, Taiwan,
May15-18, 2011. (Accepted)
[3] C. S. Hsieh, D. C. Liaw and T. H. Lin, Optimal Filtering Method to Structural
Damage Estimation Under Ground Excitation (II): Constrained Optimization
Approach, The 8th Asian Control Conference (ASCC 2011), Kaohsiung,
Taiwan, May15-18, 2011. (Accepted)
[4] T. H. Lin, S. L. Hung, T. Nagayama andY. Fujino, Study of Energy Harvesting
Technology in Structural Health Monitoring, Proceeding of Fifth World
Conference on Structural Control and Monitoring, Tokyo, Japan, July 11-13,
2010.
[5] D. C. Liaw, S. L. Hung, Y. H. Hsie, T. H. Lin and T. Y. Lu, Developing of
Adaptive Wireless Sensor Platform for Structural Health Monitoring ,
Proceeding of Fifth World Conference on Structural Control and Monitoring,
Tokyo, Japan, July 11-13, 2010.
[6] M. R. Yu, R. J. Wu, T. H. Lin and S. L. Hung, Application of Semiconductor
Type CuO-ZnO Material in Room Working Temperature for Ethanol Sensor,
Proceeding of 8th Asian Conference on Chemical Sensors, Daegu, Korea,
November 11-14,2009.
[7] S. L. Hung , T. H. LIN , and F. QIN, Developing of .NET Based Imote2 Data
Logging and Signal Processing System for Wireless Structural Health
Monitoring Sensor Networks, Proceeding of Eleventh East Asia-Pacific
Conference on Structural Engineering & Construction (EASEC-11), Taipei,
TAIWAN, November 19-21, 2008.
[8] S. L. Hung, T. H. Lin, D. C. Liaw and Y. H. Hsieh, Study of Wireless Sensor
Networks Based Structural Health Monitoring in Buildings and Civil
Infrastructures, Proceeding of CACS International Automatic Control
Conference, Tainan, Taiwan, November 21-23, 2008.
180
[9] T. H. Lin ,S. L. Hung ,and Q. Fei, ―TinyOS Tutorial and Experimental
Experience for Applying Wireless Sensor Networks in Civil Engineering‖,
Proceeding of 12th International conference on Computing in Civil and
Building Engineering, Bejin, China, October 16-18, 2008.
[10] T. H. Lin and S. L. Hung, Pilot Study of Wireless Sensor Platform for Existing
Tunnel Environment Monitoring, Proceeding of Computing in Civil
Engineering, Pittsburgh, USA, July 24-27, 2007.
[11] T. H. Lin, S. L. Hung, Development of Wireless Sensor Device for Rapid
Deploying in Buildings and Civil Infrastructures, Proceeding of The 3rd
International Conference on Structural Health Monitoring of Intelligent
Infrastructure, Vancouver, Canada November 13-16, 2007.
[12] 林子軒、洪士林、溫俊明、洪冠豪、問世賢,「基於無線感測網路之結構健康
監測系統架構與實驗分析」,98 年電子計算機於土木水利工程應用研討會
論文集,新竹,台灣,2009 年九月。
[13] 廖德誠、洪士林、林子軒、王世昌、吳立祥, 「應用於土木工程結構健康
監測之智慧型無線感應器系統平臺開發」, 系統科學與工程會議論文集,
宜蘭,台灣,2008 年 6 月。
[14] 林子軒、洪士林、溫俊明,「案例式推理(Case Based Reasoning)系統於土
木工程結構補強之應用」,台灣混凝土學會 2007 年混凝土工程研討會論文
集,台北,台灣,2007 年 11 月。
[15] 林子軒、洪士林、張瀞云,「無線感測網路應用於智慧型建築之防災系統
與網路查詢介面之開發」,96 年電子計算機於土木水利工程應用研討會論
文集,淡水,台灣,2007 年 9 月。
[16] 林子軒、洪士林、吳仁彰、陸翔寧、柴梅熙(Murthy),「無線感測網路應
用於醫院建築中麻醉氣體之感測」,96 年電子計算機於土木水利工程應用
研討會論文集,淡水,台灣,2007 年 9 月。
[17] 柴梅熙、林子軒、吳仁彰、黃裕清、陸翔寧、洪士林,“An active 433 MHz-W
UHF RF powered chip sensor for wireless monitoring of anesthetic agent
Sevoflurane”,中華民國第一屆生醫檢測理論及應用研討會,台中,台灣,
2006 年 11 月。
[18] 林子軒、洪士林,「無線感測網路平台 MICA 於土木工程監測之應用」,第
八屆結構工程研討會論文集,南投,台灣,2006 年 9 月。
[19] 黃裕清、吳仁彰、林子軒、洪士林、柴梅熙(Murthy)、陸翔寧,「導電性
高分子應用於麻醉氣體之量測與理論模擬 」,中華民國化學年會論文集,
高雄,台灣,2005 年 11 月。
[20] 林子軒、洪士林,「RFID(Radio Frequency Identification)射頻辨識於土木工
程之應用」,九十四年電子計算機於土木水利工程應用研討會論文集,台
南,台灣,2005 年 9 月。